It is the 23rd of August 2024 as I’m typing these lines from a quiet room at my grandparents' house in Hamburg. I’ve returned from the UK about two weeks ago, and I’m finally getting a bit of a break after finishing a first, incomplete version of my master’s thesis, so I have the time and leisure to write this article that I’ve been thinking about since November 2023 already. It’s been an absolutely wild year in terms of how much happened in my life, and especially things I thought about and want to talk about in detail. I’ve been thinking about what I want to write about in this article and there’s no question it’s going to be so incredibly broad that this might easily set the record for “most incoherent piece of writing” on this blog. At the same time, it’s all connected in one way or another so it only makes sense to write about it in context. Also, I’m just lazy lol. I feel like my writing would probably benefit a lot from being split into more, smaller and focused articles, but alas.

You see, I haven’t published an article in the past, checks notes, 13 months (me reading through all of this again after finishing… well make it 17 months)… and that’s certainly not because I didn’t have anything to write about but rather, because I didn’t have the time (for the most part). My memory sucks (but more about that later) and I’ve already lost the opportunity to write about some of the things that happened, or at least, I wouldn’t be able to still capture the state of mind I had during or shortly after these events. This is one of the things I’m the most afraid of, and why not being in the position to start working on this article sooner gave me a lot of anxiety - I’m afraid of forgetting things, and so I love going back to my own writing.

I’ve probably re-read the Zugspitze article at least five times since publishing it, for one thing because I’m an egocentrist who loves hearing himself talk, but also because it gives me the opportunity to relive that experience and remind myself of how I truly felt about it back then, an effect that is that much stronger because I broke with tradition for once and actually finished that article within a week of the hike. I love talking about things that interest me and sharing insights I think might be useful, but just as much, I enjoy writing for myself. In hindsight, I should have taken notes throughout this year, but I still somehow kept telling myself “I’m going to start writing that article any day now”, and so inevitably some of the vivacity of my memory is lost forever now. There are a number of exciting things I did over the last 1.5 year that could be absolutely worth talking about in some capacity, but I barely, if at all, took any notes and so it’s almost impossible to say anything interesting about them at this point. Among other things, I traveled through Japan for 3 weeks in June 2023, and I already started an article about that, barely covering the first 4 days of the trip in 8000 words. But I wrote all of that 11 months ago, and I’m not able to continue that with a similar level of detail and liveliness. You’ll still get to see this article in some form some day though. Further, I went to the Chaos Computer Camp in August 2023, and traveled to Portugal with friends shortly after that, the first time I did a proper trip with friends I think. Both events worth talking about, but I probably never will.

The article is so massively long that I cannot in good consciousness recommend anyone read all of it, also because it is so over the place that a large portion will likely not be that interesting to any given reader. So, before we get into it, here’s a quick overview of the contents of this post:

In Section 1, I talk a little about my background and how I ended up at Oxford, including the application process and how I got my accommodation. In Section 2, I go through the first two or three weeks after arriving in Oxford, including how I moved in and made my first friends. Section 3 is all about the actual studying experience: the computer science department of Oxford, a detailed description of the courses I took and the general studying experience there. If you’re not interested in Computer Science, then you can skip almost the entire section, except perhaps Section 3.7 where I rant about the grading system in the UK. In Section 4 I go through my specific accommodation, the city of Oxford as a whole and some general notes on life in the UK. I’ve dedicated Section 5 in its entirety to talking about food. Take a look if you needed a few more reasons to convince you to not go to the UK. In Section 6 I explain what functions the college and MCR serve, but especially how much I liked my college, the events and the people there. This is mostly a look into what my life looked like in Oxford. Building on that, in Section 7 I talk about all the other things and activities I did in Oxford, aside from my engagement in the college. Section 8 covers a lot of the things I take away from my year there, including a lot of very personal learnings and introspective ramblings. Section 9 covers my experience looking for jobs, which I did on the side while studying in Oxford, including some funny and less funny recruitment processes I encountered on the way.

Alright, enough of the ramblings, so let’s get into it. I have no idea where I should start and this article will be all over the place anyway, but let’s talk about the facts and some context first: how I got to studying Computer Science at Oxford.

The Road to this Point

Like the vast majority of people after finishing highschool, I didn’t have much of an idea about what I wanted to do in life. I’ve been going to school for two thirds of my life, now what? I somehow knew I wanted to do software-stuff. I don’t remember how I knew, but I’d grown up with computers, did a little bit of programming in my free time and in my first school internship, and in terms of job prospects, you can’t go wrong with this. I took the path of least resistance, and went for studying computer science for my Bachelor. Having just finished school, that was the most obvious choice and the farthest away from actually making a choice in hindsight: just continue studying! For context, university studies in Germany are incredibly cheap compared to most other countries, for a “tuition fee” of about 300€ per semester.

At the same time, I do have ambitions, I had pretty good grades and I was vaguely aware that there was such a thing as “prestige” and university rankings. So I checked said rankings online, looked at the top universities in Germany for computer science and filtered it down to RWTH in Aachen, KIT in Karlsruhe and TUM in Munich. The latter was kinda out of the question because Munich is expensive, and between KIT and RWTH I vibed more with RWTH, so not much later I packed up my stuff to live and study in Aachen for 3 years. Fast forward 2.5 years, I knew I was going to have a Bachelor soon but still had no real idea what I wanted in life. The CS Bachelor at RWTH (and most CS courses in general) tend to be pretty broad in my experience, which isn’t a bad thing, but means that you get little opportunity to specialize much in anything and decide what you really want to do. CS is a big field after all, so most of my undergrad friends similarly had no idea what they wanted to do next. I felt that, most importantly, I was lacking the practical industry experience to know what kind of work I would enjoy. So I ended up going for internships first, a 3 month one at a startup in Munich and another 6 month one at a big German tech company. Both experiences were really helpful in shaping my ideas of what work was like, but they didn’t help me decide what I wanted to go for either.

At some point I decided I was going to do a Master’s, though I was very unsure if this was what I needed, compared to getting FTE work experience. I’d had a strong GPA from my Bachelor and this time I was thinking bigger: I’d searched for the best CS undergrad universities in Germany, but now I was going to look globally. I had a relatively clear idea that I didn’t want to go to the USA, if nothing else, because of the insane tuition fees. Several others were filtered out due to tuition fees, including e.g. Singapore. I’d also looked at Japanese universities, but found somewhat disappointingly that they were all very research-focused and I was more interested in a taught course. I ended up applying to both Tsinghua university in Beijing and Oxford in the UK, while keeping the TUM in Munich as a backup plan. With some surprise, I was accepted - by both universities!

Now, to some extent, judging what a university can offer you is difficult from a distance, and a significant part of this exercise was about the name: I wanted a Master’s degree from one of these universities so I could put their name on my resume afterwards, because it would open doors for me that might otherwise be closed for the rest of my life. In that regard, Oxford was simply the more established name and thus the better choice over Tsinghua, though there’s no doubt that could have been an incredible experience as well. Add to that the courselength of only 1 year at Oxford, which suited me well since I was a bit unsure whether I wanted to study in the first place. On the other hand, Oxford had (by German standards) insane tuition fees of about 37.000€ compared to 15.000€ in Tsinghua, and I initially (before I got accepted) said I was only going to accept Oxford’s offer if I got a scholarship. The CS departemnt claimed that 80% of its students were studying on a scholarship, and the application process made it easy to apply for them: just tick a box and you’ll be considered for almost all of them automatically. But most scholarships don’t give you a decision until after the deadline for accepting the place, so you essentially have to promise that you’re able to go self-funded either way. I eventually decided to accept the Oxford offer, thanks to my brother who lent me the majority of the tuition fees. On a side-note, UK nationals pay only about 15.000€ of tuition fees (thanks Brexit) compared to us poor internationals. That seems wild to me when I compare that to my undergrad, where internationals even sometimes got tuition fees slashed, rather than having to pay more than twice as much… the uni was trying to milk us for money, and this will be a recurring theme here. Another note on scholarships: based off of the number of people I’ve talked to, I would say the scholarship-ratio in my department was roughly 20% - the inverse of what they claimed.

Let’s talk about the application process for a bit, because some of that was interesting. For Oxford, you had to have good grades from undergrad: their website listed a minimum GPA of 1.4 for German students, though ultimately there were people on my course whose grades were slightly worse than that. Further, you needed to write a “motivational letter”, which until this day I see as nothing but an anxiety-inducing exercise in bullshitting where you try to craft a narrative about yourself. It’s not that I don’t think there’s some value in these, but ultimately these written letters might as well be AI-generated and I feel like no one would notice a difference. Additionally, you needed three reference letters from professors. That was a requirement at every international university I looked at, and it’s absolutely horrible design. The RWTH in particular was known for having one of the highest students-per-professor ratios, and I suspect the only academic who might recognize my face and have anything of value to say about me, was my thesis supervisor. I got another professor to write a reference letter for me based on me getting a good grade in one of his courses, and it said nothing meaningful beyond “guy took my course and got a good grade” which is like… they could tell that much from my transcript. Not like I’m blaming the guy, but it calls into question the point of the entire exercise. The worst is, there’s certainly a low chance of getting accepted into Oxford, to the point that the university explicitly recommends applying to other places as well. I think that’s a good recommendation, but also, you’re making this really hard. I would need three reference letters for every single application, and every one of these universities has their own form with its own structure etc. to ensure you cannot use a non-specific reference letter for applications at multiple universities. In other words, if I wanted to apply to $n$ universities, I would need to bother an academic to write a reference letter for me a total of $3n$ times. You know how reliable academics are - this is a horrible exercise and essentially means that you couldn’t realistically apply to more than two or three universities.

Anyway, I still managed to get that sorted out obviously. I also needed a proof of my language ability, which meant doing an English test in my case. I went with TOEFL, and just barely passed the requirements, with 112/120 points, where 110 were required by Oxford. I got the full 30 points in reading and listening, followed by 27 in writing and 25 in speaking. Perhaps this should have been some sort of wakeup-call, that the standards were really high. It didn’t lead to any problems, but I only realized once I was there, that my English was not as great as I thought.

The final step was then an interview with a professor. I later found out that these were carried out randomly - a significant number of candidates were accepted without an interview. In any case, my interview was with Prof. Ismail Ceylan - a cool guy and a fun interview honestly. He mostly asked me about fundamentals of complexitiy theory, graph algorithms etc. I don’t remember all the details, but the first question he asked was: what grows faster? $2^n$ or $n!$ ? This question nicely set the tone for the interview in my opinion: I first answered $2^n$ from gut feeling honestly. He then asked me for an explanation, I started actually thinking about it and quickly realized that this wasn’t as difficult as I first thought, and that the correct answer was $n!$. Overall, this was a great interview format in my opinion: you talk to someone who knows their stuff and asks you difficult questions, asks you for justifications, gives you hints if you need them or if you forgot a random formula or something etc. Something very interactive. We will talk about job search later, and this interview right here is a great counterexample to all the ways to how it’s not done that I encountered then. On a sidenote, I happened to study logaritmic laws not too long before the interview and it came in handy. Cheers to my propensity to create random flashcards for things I deem useful.

So, I got accepted some time after that, yay! I was quite happy to say the least. The next thing was paying tuition fees, which was a weirdly convoluted process: I first had to provide a document proving my ability to pay, which was surprisingly difficult to procure. It took me about a month of back-and-forth with the university and my bank to get this sorted out in the right format. After I submitted that, they sent me the payment details right away and I paid within a couple days. So what did we need this for? Why couldn’t I pay right away? Ugh. Anway.

Another step was figuring out accommodation for my time there. By the way, I got assigned a college, St. Anne’s college in my case, though I didn’t quite know what that meant. Explaining what a college is, is a bit difficult but I hope I can get the spirit across over the course of this post. In any case, I had the option to ask the college if they or the university themselves could provide me with accommodation, or look for private accommodation. I asked college for the options of course, and they sent me two - one of them university accommodation, and one of them provided by college. I preferred the college one slightly, and it said 700£ per month in the current year. That “current year” part is very important - but I didn’t know that back then. In any case, I looked for private accommodations as well, but found them to have similar rent, and because it would be easier to get college accommodation rather than organize something myself in a foreign country, I went with that. To make the choice binding, I sent them my deposit, and a month after that I got the rental agreement - suddenly stating 900£ per month and a few weeks longer contract, leading to an increase in rent by fourty percent for the full year, compared to what they initially told me. And because they’d had us confirm our interest, pay the deposit and then wait for a while, it was now impossible to look for private accommodation which had suddenly become significantly cheaper in comparison. At this point you feel scammed, and there’s nothing you can do. The original 700£ was already expensive, but unfortunately Oxford is just expensive. But now, I was signing up to pay rent more than four times as much as I paid during my time in a 15m² room in a student dorm in undergrad. This was another punch in the stomach of my financial planning for the year, after realizing that I was not getting a scholarship.

I knew I had some monthly support because my parents sent me the governmental child allowance, and some from my brother as well. I also applied to the German government’s international studies support, which is fortunately much more generous than the national studies support. Most of it was scaling with tuition fees, and the 37.000€ easily surpassed the 5.000€ they set as the maximum to get the highest possible support. So I was granted 850€ of monthly government support, which is very generous and not even enough to cover my rent. Heh. I asked my brother if he could increase his monthly support which he kindly agreed to, so I was ultimately living on a budget of 1700€ a month. For reference, this is noticably more than what I earned during both my internships (after tax), during which I felt like I lived like a king, but in Oxford this was barely enough to get by! I had to borrow another notable sum from my brother as a one-off thing, because they then asked for 3.5 months of rent in advance - ouch. And even after that, I still distinctly remember having a balance of 30€ on my bank account on the 25th of October, at the end of my first month right before my next support payments hit. After that, things were going up again, but this was uncomfortable.

I will have to say here that I am in the incredibly lucky position to be born in Germany, a country that provides such abundant support, have supportive parents as well, and most importantly, an incredibly generous brother who is also wealthy enough to promise me to “help out” if I ever needed it. But I don’t want to imagine what someone less lucky in a similar situation as me would have to go through - someone who can’t just go “hey bro can you send me a thousand Euros?” but has to watch in dread as their account balance keeps ticking down with every passing day. Oxford’s posing about their openness, saying stuff like “we’re proud to make education accessible to everyone” makes me feel a bit sick. Looking at all my friends in Germany, I’m pretty sure not a single one of them would have been able to accept an offer of a place at Oxford, for purely financial reasons. Capitalism makes me mad, and I say that as someone who’s in the lucky position to not have to deal with that shit in a serious way most of the time.

Whew, okay so let’s get on with the start of my stay now.

The First Weeks

I arrived in Oxford roughly a week before induction began. I walked to college with my big suitcase and registered myself, picked up my student ID and keys and began the long walk to my student accommodation: Robert Saunders House in Summertown. My college is already located somewhat north of city centre: not that far, but Oxford is a small city and most of my friends from other colleges considered St. Anne’s to be pretty far away. Summertown was even farther north, about 2km from college or 3km from city centre. I passed many classically-Oxford old buildings, which are genereally reminiscent of medieval castles, but like, smaller… I got a bit anxious on the way as I passed nothing but residential houses on the way, fearing that I would have to go 3km to city center to buy groceries. But fortunately, Summertown just so happened to be a sort of suburb, a lively little place with a small core of supermarkets, shops and restaurants surrounded by a cozy residential area. My student accommodation was okay, as long as you do not take the price into consideration. I was on a hallway with 4 rooms total and a shared kitchen, shared bathroom and my own room of 20m².

My first evening in Oxford I spent exploring the local supermarkets for a bit, thinking about all the things I needed to buy for my room and for the kitchen, looking online for a used bicycle and planning some socializing. I had joined a group chat for new students at my college, and someone there happened to ask if people wanted to meet in a pub to get to know each other or something. If you know me, you know that going to a pub and meeting a bunch of new people all at once would be a bit much for introverted little me. But I wanted to make the most of my time here, and I had a couple of days with little to do anyway, so I messaged the guy who asked and signed myself up for this. I got up early the next morning and procured my used bicycle from the other end of the city for 70£. This significantly shortened my commutes to the city to just 10 minutes, rather than about 30mins of walking or paying for the bus every time.

Later that evening I went to a pub and met the others who’d signed up for the pub evening organized by that guy in the group chat. I was a bit anxious about this and not very motivated at first, but it turned into a really nice evening. There were about 10 people at the pub, all of then new master’s students who were starting at the same time as me. All of them were new to the city, open to socialize and generally nice people. I didn’t manage to talk to every single one that evening due to numbers and seating logistics, but certainly made some connections that day already. At least half of them I kept seeing, whether by chance or on purpose, throughout the following months in Oxford. Since these were all people from my college, and colleges are not limited to certain subjects, it was a very diverse group of people from all over the planet and all over the academic spectrum. I spent a lot of time chatting with an American studying economic history, and a Chinese-Italian doing education. The topic of the college dining hall came up, and we decided to go there together for lunch the next day.

The next morning, I realized I didn’t have anyone’s phone number, so I just asked in the big group chat where our organizer had asked about the pub evening, if someone wanted to join me for lunch. This worked out as planned, and the two I’d chatted with the previous day showed up. Interestingly, an Indian guy studying International Relations also joined us - it turned he was one of my three flatmates, and he got my number from another flatmate I’d met the day before. Sometimes things work out nicely. We became good friends after this. It was still three days until the start of the induction week, and I felt like I’d already made a few friends.

Fast forward to the induction week: this was the “0th” week of my studies, a mixture of events organized by my college, the MCR of my college and by my department, to give us opportunities to get to know all of them, get some bureaucracy sorted out, meet people and make friends. A short explanation for now on what college and MCR are: colleges are (relatively) nebulous entities and I still couldn’t tell you entirely what their exact role is. But generally speaking, every student is associated with a college, and the college is responsible for all the administrative stuff that is not very clearly the jurisdiction of your department. I registered as a student with my college and got my student ID from them, my accommodation is managed by my college, my university email is managed by my college. The college has its own premises with buildings and facilities, so if I want to eat uni-subsidized lunch, I go to the college dining hall. If I am in financial or psychological trouble, I can contact the college advisory or college nurse, and even if I want to ask for a project extension for one of my computer science course projects, I discuss this with college. On a sidenote, there are only three universities using the collegiate system in the UK, and it has nothing to do with the concept of colleges in the USA.

The MCR is an institution that I am part of: the Middle Common Room, which is comprised of all the college’s Master and doctoral students. There’s also the Junior Common Room (JCR), the equivalent institution for undergraduate, Bachelor students. The term also refers to an actual room: both institutions receive their own room they can sort of do whatever with, usually serving as a hangout space and a room where events and parties are organized. The JCR and MCR distinction is important because they tend to have very different vibes, are treated differently by college and run by separate committees, but more on that later.

So, induction week: we listened to a speech by our college principal, were shown around college on a tour, got a lot of info dumping in the department, had welcome drinks & pizzas and fun exercises to get to know other people. These first five “official” days were very important, at the very least, in building my social circle and I got to know a lot of people there that accompanied me to the end, that I have become great friends with. But to be honest, I don’t remember many details from that week so I can’t say too much here. I don’t even remember how exactly I met some of my now closest friends in that week.

Another important part of the week was the university’s “Freshers' fair”, where all the university’s societies and other institutions got the opportunity to introduce themselves. They had an entire massive building booked out for this, which was necessary because there was so much. I probably spent three hours at the fair. “Society” is a fancy word for some sort of university club that people join for fun, and Oxford has everything you could possibly think of. There’s a society for almost every country, one for every possible type of sport, one for video games and one for pen & paper RPGs, one for anime, a Taylor Swift society, at least seven choirs etc. There’s even a caving society, as in, people who explore caves. All of these are pretty much entirely student-run. I signed up to at least ten mailing lists that day, and was looking forward to try a lot of different things. In the end, I stuck with the Japanese society and one of the non-auditioning choirs. This turned out to be easily enough, and I would have loved to do more: I signed up to mailing lists for the German society, the Track & Field society, the Taylor Swift society (heh), the robotics society, the computer science society, and several others that I already forgot about.

That was my first week in rough terms, and after that we finally began studying. I’ll talk about that in detail in the next section.

Master Studies

I’d looked at the course selection at Oxford before applying of course, but I didn’t have a very clear picture of what I wanted to study until I applied. In a way, I went down the path of least resistance again: I decided to focus on AI. In many ways, that was an obvious choice: the field is certainly in demand and there are a lot of interesting developments happening there over the past couple years. There are many things I want to learn more about, but not all of them can be studied at university, but this was one of those you could study. I had some basic interest in AI, and Oxford had a good course selection on offer, so this was settled relatively quickly. In fact, AI was the only possible “subject area” that Oxford offered enough courses for to fill out the entire year. On the one-year CS course, I had to complete at least 6 courses plus a Master’s thesis, but I was completely free to choose which courses exactly I wanted to do. There were 7 courses that were clearly focused on AI, but if you wanted to study any other field, you could get at most 3 courses that were meaningfully related. Since there are no compulsory courses, you could technically come up with the most pointless ragtag mix of subjects one could imagine and go out of this degree having learned a bit of everything but nothing comprehensively, which seems… a bit strange? Anyway, I went with 6 of the 7 AI courses, dropping only the AI+Bio course. I had three courses in first term, three in second term, and the thesis work plus exams in third term. An overview of my courses:

Machine Learning

This is the department’s introductory machine learning course, covering a bit of everything. But while we did discuss a wide range of architectures, the course was very much focused on the theory behind stuff. If you know me, this is already a bad start, but this one honestly made me feel like they should have just named it “advanced probability theory” or similar. Worst of all, the theory parts that we did talk about are mostly theoretical frameworks that are entirely useless for anything you’ll ever do in the real world. I mean, one of the most difficult topics that I feel like we spent a lot of time on, were Maximum Likelihood estimators (MLE) and Maximum A-Posteriori (MAP) estimators. A very, very abridged version of what this does is to explain why it is valid to e.g. estimate the mean of a statistical distribution by taking the average over a bunch of samples from said distribution. This may seem obvious, but it is not always. The sad thing is that this is only really relevant to basic regression. In general, we talked a lot about basic regression, Support-Vector Machines and the likes, rather than, you know, the Neural Network stuff that almost everyone came for. The lecturer started his second lecture with a screenshot of a Facebook post:

I would say I generally look favorably upon those defending the mathematical, theoretical foundations of anything against the masses of those who want “just the practical stuff”, but this is different. Machine Learning in general has become a massive field and theory is completely useless to explain a lot of its facets!

Computer Science, as a subject area like you study in university, is much more theoretical, and much more rigorously mathematical, than you would think if you didn’t study it. You don’t just learn programming in a CS course, this is perhaps 10% of a typical CS course, and I was spooked by this as well: when I started my undergrad in CS, I couldn’t believe how much time we spent on dealing with theoretical maths, and it was genuinely hard. I say that as someone who loved maths in highschool and was considering studying mathematics. But in contrast to some of the more foundational areas of computer science, ML is highly empirical, and this has become a recurring theme throughout my courses. A lot of them focus on theory, teaching you theoretical knowledge that is either a) useless in practice, or b) shifts the focus to only those facets of ML that can be explained by theory. This is not only a problem with the courses I was taught in Oxford, but with the field as a whole. It feels like a large fraction of machine learning researchers are mathematicians who are desperately trying to make it seem like their background is useful for this field.

Don’t get me wrong: rigorous mathematics and proofs have their place in ML. Some incredible insights in the field have come out of purely theoretical derivations. But this is a field where the stupendously simple technique of blacking out parts of an image, and then letting your AI model try to fill in these parts as pre-training, has produced one of the most successful and influential papers in the field. It just works and no one fully understands why. If you introduce a simple idea (make a small, intuitive change to an architecture) and show that this leads to improvements on typical datasets in a concise 3-page paper, then there is very little value in spending another 8 pages to prove that “this is valid” from a theoretical point of view.

Coming back to the Facebook post above: I do not expect a basic ML course to be all about “learning PyTorch” or something similar. But if you decide to focus on theory or mathematics, you deliberately ignore the >50% of the field that are entirely empirical, but still useful or important. I would have been very much on-board with the original post if the example he gave was “pip install openai and then start calling an LLM API”. But as it stands, my only reaction to that can be… “yes that is (also) what I want to learn!”.

“training a giant convnet on multiple GPUs” is not a simple task, and you could start with interesting questions such as “how do you train a model that doesn’t fit on a single GPU?”. I was never given an opportunity to seriously consider this question, until I was asked exactly that two months ago in a job interview. As it stands, the post has the same vibe to me as a mathematician saying “we don’t deal with numbers” as he gives you an arrogant look. My prof echoing this statement makes him seem out of touch with both the reality of the field beyond pure research, and his target audience.

To go full circle: my ML course had 20 lectures. In lecture 14, we talked about neural networks for the first time. Enough said.

Graph Representation Learning

GRL might have been the most interesting course I took, and it was held by the professor who’d interviewed me for my application to Oxford. Perhaps the most well-known subareas of ML, by application/modality, are natural language processing (NLP, usually written text) and computer vision (images/video). My impression is that almost no one has heard of graph learning, which is ironic, consdering that Oxford offers a course on it, but not on NLP. In any case, it is still a highly relevant area whose usefulness is often underestimated.

Graph learning is all about machine learning applied to graph data, where “graph” refers to those structures consisting of nodes and edges. This might seem a bit niche at first, but a surprising number of interesting types of data can be formulated as graphs. Perhaps the most relevant in terms of research today, are molecules. These can be represented as nodes (atoms) with edges (bonds) connecting some of them. In fact, graph learning has already been in use for a while, to identify interesting chemical compounds for medicine discovery and many similar tasks. A classic example is the ZINC dataset, which contains graph representations of molecules with the goal to predict their solubility. Many other problems can be framed as graph learning problems, such as supply chain optimization.

This was a great course that managed to strike a strong balance between theory and practicality. Not much more to say here.

Computational Learning Theory

CLT was an… interesting… course. This is the last one of the three courses I chose for the first term. It was perhaps the hardest course I ever took in my life, and choosing it had a significant negative impact on both my final grade and my mental health.

This course essentially introduces a fundamental theoretical framework for machine learning. This goes far deeper than the theory I mentioned in the context of the ML course: CLT asks what tasks can even be learnt? I’ll try to give a basic introduction and example here because it’s difficult to grasp what this course is about otherwise, feel free to skip four paragraphs ahead if you don’t care.

An essential concept to explain here is PAC learnability: a problem is PAC-learnable if you can give an algorithm that produces a hypothesis with an average error below some value $a$, getting error ratios above that at most $b$ percent of the time the algorithm is run, if you feed it with some number of example datapoints $p$ such that $p$ is polynomial in $\frac{1}{a}$ and $\frac{1}{b}$, allowing something like “if you want half the error/failure ratio, you can have twice as much training data”. PAC stands for “probably approximately correct”. “probably” because the algorithm only has to work $100-b$ percent of the time, and “approximately” because its answer is allowed to be, on average, only $a$ percent wrong.

The most basic example here is the rectangle learning game. Imagine you have a square playing field of size 1x1. There is an axis-aligned rectangle in that area, with all points inside that rectangle getting label $1$ and all points outside getting label $0$. Given a bunch of points with x/y coordinates as well as their label ($0$/$1$), with probability $100-b$ your algorithm should produce a rectangle that is at least $100-a$ percent correct, compared to the actual rectangle.

The most intuitive algorithm here actually works: start with an output hypothesis that is the full 1x1 square, and whenever you read an example point with label $0$ (i.e. outside the actual rectangle) you shrink your rectangle hypothesis until that example point is just ouside it. Showing how many examples you need to achieve the $a,b$ bounds is not trivial though.

The easiest way to do this is to imagine you have an “error budget” $a$, which you distribute across all four sides of the rectangle: a rectangular buffer zone on each side of the rectangle that is exactly wide enough that you’d expect a fraction of $\frac{a}{4}$ of all randomly chosen points to be in each of these strips. Now you need your $p$ examples to contain at least one point in each of these buffer zones. If that happens, your algorithm will shrink your rectangle to the point that it is at most by a factor of $4\cdot \frac{a}{4} = a$ bigger than the actual rectangle, for an error ratio of at most $a$. Your algorithm can only fail if at least one of these 4 buffer zones doesn’t get any example point. The chance for any point to be in a specific one of these four buffer zones is $\frac{a}{4}$ by definition, so the chance that, out of $p$ points, none of them is in that buffer zone, is $(1-\frac{a}{4})^p$. The chance that any one of the 4 buffer zones doesn’t get a single point is at most $4\cdot (1-\frac{a}{4})^p$. This is an upper bound on our failure probability $b$. Now you set those two equal, solve for $p$ and get $\frac{4}{a}\log\frac{4}{b} \geq p$. We have an upper bound for the number of training examples our algorithm needs to achieve failure chance $b$ and error ratio $a$, and it is indeed polynomial in $\frac{1}{a}$ and $\frac{1}{b}$.

This course reminded me a lot of my Mathematical Logic course back in Aachen during undergrad, which had been the hardest course for me until CLT. Both of them were incredibly challenging at first, but in the end you leave with the feeling that you’ve gotten really good at the kinds of exercises that are typical for these subjects, and both of them are a lot of fun to think about. They’re not at all useful for me :D While the idea of being able to say that certain tasks are not effectively learnable with any reasonable amounts of training data is very interesting, real-world tasks such as, say, machine translation, are way too complex to capture them in PAC learning framework. Besides, we got to the point where I feel like “exponential training data requirements” are not going to stop people from trying.

An estimated 60 or so people started the course together with me in week 1, a similar amount as you saw in the ML or GRL courses, as I assume everyone in “the AI student crowd” wanted to take this course. But the lines quickly thinned, and I’m pretty sure that less than 10 people ended up doing the final graded project in CLT. I can’t blame them: the course immediately made it clear that it was hard theory and not much else, and more than anything, the very first homework we got in week 1 was absolutely brutal. It consisted of 4 exercises, on which I spent about 40 hours in total. I got 1 and 3 solved, and gave up on 2 and 4 after many hours. This homework almost single-handedly wrecked my confidence and made me believe I was in the wrong place. By the middle of week 2, I was seriously consdering quitting my degree and going back to Germany because I thought I just wasn’t smart enough. I still made it through because I never quit, I’d done some communal crying with my fellow CLT victims after lectures so I knew it wasn’t only me, I’d already made some wonderful friends who distracted me from CLT depression, and I seriously lowered my expectations: this wasn’t highschool or Aachen, and I couldn’t expect to get close to a perfect score in everything I did anymore. More on this later though.

Life got easier after that first homework: my first term was still unholy stressful, but none of the following homeworks in CLT cost me more than 12h after that. Whew. They really tried to scare people away with that first sheet. I remember writing on the end-of-term feedback sheet that the first homework had a significant negative impact on my mental health. Honestly, this course may not have been worth taking. But at the very least, I got the street cred. One notable experience during induction week was that you’d talk to a lot of people, I would say roughly half of those who began their CS Master’s with me, and you’d ask each other what your research focus is. The majority of people said “Machine Learning” and we’d give each other the “one of us” look, or they did something different, and lowkey rolled their eyes at me as “just another one of the boring ML people”. I can’t judge them for that to be honest. But anyway, after I finished CLT, I was told that I’d earned my place as one of “the good ML guys”. Heh. We did a “CLT survivors' party” afterwards, which was great.

Computer Vision

Entering the second term courses: this was a great lecture. It was held for the first time apparently, and the guy doing it was a bit chaotic, having some errors in his lectures and exercises and changing things at the last minute all the time since it was his first one, but overall, he did a great job. Easy to digest, well-structured, well-explained, really helpful professor, and fun to follow along with. He covered basic image processing stuff and some classic computer vision, and then focused on the neural networking stuff. Not much else to say here. This is the only course that actually taught transformers and attention, the fundamental building blocks of most current state-of-the-art models such as LLMs.

Geometric Deep Learning

This is essentially a generalization of graph learning. In geometric learning, you look at learning models or paradigms as an attempt at capturing certain kinds of invariances. As an example, most computer vision models try to capture shift invariance, i.e. invariance to movement. If, for example, you build a classifier that recognizes cats, you would want it to output “cat” regardless of whether the cat is in the center or top-right corner of the image. In other words, sliding the image around doesn’t change the result.

Graph learning captures another kind of invariance, the permutation invariance. Typical graph learning models follow the message passing paradigm, where every node in the graph updates its state as a function of a) its own state and b) all of its neighbors' states. The crux here is that there is no ordering among neighbors: all of them are treated the same. This makes the entire architecture permutation invariant: if you feed in the exact same graph in terms of nodes, their features and connections, only in a different order, the result should still be the same.

After a way too long introduction, this is the core of geometric deep learning: embedding this idea of different architectures capturing different kinds of invariances into a mathematical model and some sort of hierarchy. This keeps going until you reach manifolds: another domain of objects (similar to graphs) which have their own type of invariance that we are interested in. Except that all of this is highly theoretical and has no known practical use, as one of the teaching assistants told me. Ultimately, we really only talked about graph learning some more in this course. The idea of thinking about image, graph and other domains in terms of object classes and invariant functions over them is certainly interesting, but I don’t feel like I got a lot out of this course. I’m struggling to recall what we spent 20 lectures on here. I just checked my lecture notes again and realized that the last 4 of those were guest lectures, and I wrote down a total of 15 words on the last 3 lectures before that.

Uncertainty in Deep Learning

This was the most chaotic train wreck of a course I’ve ever seen. The basic idea of this course is that we’re interested in the uncertainty of a machine learning model, i.e. we not only want it to produce answers to our questions, but also add confidence scores so we know how sure it is. This is super interesting and useful.

That said, the course was not. First of all, the lecturer held it for the first time and utterly miscalculated how long he would take for the course content, considering his teaching style. We managed to go through about half the course content in the 18 lectures he held. He took things reeaaaally slow: as a first taste, in the second lecture, he did an experiment with us. Now let me preface this by saying that I’m all for engaging your audience, but… not like this. Effectively, he gave us 5 minutes to discuss with the person next to us about what we think the probability of it raining tomorrow is. Then we collected people’s results and discussed them. Cool. Then we did the exact same thing for 4 more questions. And he repeated a similar experiment in the next lecture. This was very tiring.

Once he was done giving us an idea of the content of the course with way too drawn-out experiments, we jumped straight into theory. That mostly involved some difficult probability theory stuff, and he insisted to prove every single thing he did in exhausting detail on the whiteboard in front, and most of the time he’d even give us 10 minutes to try and solve it ourselves first, before going through it in detail himself, or letting one of us present their proof. Again, I am not entirely opposed to this: I learned a lot about probability theory and some very interesting derivations this way. They were genuinely hard, but I was able to follow along nicely because he tried hard to make us understand the stuff. But we completely lost track of what we were doing and why, and never got around to the interesting things.

We learned in the practicals that we can get uncertainty estimates by making the parameters in our neural networks not simple weights, but rather normal distributions: i.e. a mean and variance parameter, instead of a weight parameter, for every learnable weight. During inference, you then infer a hundred times or something with actual weights sampled from those weight distributions, and then you compute a confidence score by seeing how many of your randomized evaluations came down to the same result. This idea is really cool and super simple, and the entire lecture was only about proving why it’s valid to do this. That’s a prime example of the stuff I’d discussed earlier in the section about the ML course.

The course had a few interesting things to say, but barely managed to do so while imploding in an absolute disaster of a teaching experience.

Examinations

I wanted to talk about both my experience with graded projects and exams here, and especially about the grading system in the UK. It sucks.

To give some context about my own experiences and thus expectations, let me tell you a bit about how grading worked during my undergrad: evaluations were usually given on a scale of 0 to 100 percent of points attainable, and grades were directly derived from those percentages. At 50% you would get the lowest possible passing grade 4.0, and at 100% you’d get the best grade of 1.0. Between that, grades moved by a third for every 5%, so 50-55% was 4.0, 55-60% was 3.7, 60-65% was 3.3, etc. up to 95-100% which was 1.0. This system wasn’t set in stone though and examiners could deviate from this however they wanted. This was usually only the case for exams that turned out particularly badly (e.g. the theoretical maths stuff), in which case a common approach was to up everyone’s grades, for example, by setting the pass mark at 40% and then “widen” every grade step from 5% to 6%.

The vast majority of courses was graded via exams, and within 3 weeks after grades were released, the examiners had to hold a sort of in-person consultation hour, called inspection, where students got the opportunity to look at how exactly their exam was graded, what they were deducted points for etc., and were often provided with model solutions to compare. Examiners were present to be asked any question, and could give students additional points and even up their grade if the student found a grading mistake or could argue sufficiently well that their answer deserved more points. As you can imagine, this usually turned into multi-hour discussions and negotiations by desperate students, and I’m sure examiners would have loved to not have to hold this event. But it absolutely did wonders for transparency and fair grading. For one thing, mistakes happen frequently during grading. And that’s okay, but there’s no other way to check for this. It helps to accept bad grades and to understand what your mistakes were. You don’t just try to get a better grades at an inspection, you come out of that having learned more about the subject, as well as your own weaknesses in preparing for this exam. Keep all of this in mind as we talk about the UK grading system.

The UK grading system also uses a percentage scale, where 50% is the lowest possible pass grade. There are no real “grades” derived from these percentages though, only three different classifications: the regular one, as well as “merit” and “distinction”. You get the regular one with a grade between 50% and 65%, a merit for 65-70% and a distinction for 70% and upwards. But wait, aren’t those ranges… strange? Why is the second classification only 5 percentage points “wide”? So, I’m not sure how much of this is Oxford, and how much of it is the UK at large (there are unis in the UK where you get a merit for the 60-70% range) but the idea is roughly, at least in a research or similar context, that students normally get grades in the 50 to 75, maybe 80, range and anything above that is straight up worth publishing in an academic journal or similar. As you can imagine, the effective grade range is thus compressed from 0 to 100 down to 50 to 80. Oxford grading gets extremely generous as you approach the 50% (in downwards direction) and the only time I’ve ever heard of a grade less than 50% was from someone who struggled to multiply 3 small, sparse matrices together. On the other end, scoring above 70% is already seen as pretty strong, and 80% and above is so good that I assume half the graded projects I completed didn’t even give out a single grade that high. In general, the “difficulty increase” if you want to go from 65% to 70% is reasonable, while the increase from 70% to 75% is massive, and might sometimes be completely insurmountable.

There’s a lot to say about this, so let me start with the lukewarm takes: compressing the range like this seems just… weird and unintuitive at first, but is not exactly a problem in and of itself. You could map the 50% to 75% range to a 50% to 100% range and the resulting grading system would fundamentally be very similar to the German one that I’m used to.

The problems lie elsewhere, with perhaps the “spirit” of the system, or the guidance given to examiners. In natural sciences, I would say there are two types of exercises: the clearly defined, concrete and/or theoretical ones, where there is often only one solution (or very few) and the best that someone can possibly do is to just get it right. An example of this would be any mathematical proof exercise. If your proof is correct, there is almost no way to not give you the full number of points, while any oversights or mistakes will be deducted from your points, depending on how much they affect the plausibility of the final result.

On the other hand, we have more open exercises, such as those beginning with words like “Explain […]”, or on a bigger scale, the “do some research on a topic related to this course” types of exercises that were very common during my studies. These are the only kinds of exercises in most non-STEM courses, I assume. I am not a big fan of these because I like the reliability of my “either you’re right, or you’re wrong” theoretical maths exercises, but there’s nothing inherently wrong with these. The issue here is, they don’t mix well. The mathematical problems tend to have a very clearly defined upper limit. This doesn’t hold for the second kind. There might not exist a ceiling at all for how good or interesting your research could possibly be. For the latter, you’ll have to reserve the upper parts of the range for those truly exceptional, once in a year (if at all) works to value them appropriately, and Oxford is very focused on this.

There are several ways you could combine these two kinds of exercises in the context of a fixed 0 to 100% scale. One option would be to ignore their relationship completely, meaning e.g. you rate someone successfully completing an exercise of the first kind with 100% of the points, while someone solving an exercise of the second kind really well would only get, say, 80%. Unless you clearly identify these types of exercises and somehow treat them separately, the problems are obvious: students are forced to choose courses in a way that minimizes their exposure to the second kind in order to maximize their grades.

The next option is to set the 100% bar for the second kind to something lower, so that a “really well done” submission is enough to get a perfect score. The only downside of this system is that you lose the ability to “reward” exceptionally strong works appropriately. This is what the German system I’m used to does.

The final option is to somehow force down the grades of the first type of exercise. All ways of doing this suck to be honest. Because this is exactly the route Oxford chose, I’ll bring up a couple examples as I discuss the exams I had.

In CS, we had two kinds of graded exercises as well as the thesis work. Every course was evaluated via either an exam, or a mini project. Exams were written collectively in a roughly 3 week span during the thesis phase towards the end of the course. Mini projects were done during the “break” between terms and could be about literally anything, although most of them were simply research projects where you had to write a paper. For those, the “grades above 80 are for submissions worth publishing”-type of thinking became very apparent, as it seemingly inspired the entire grading scheme for them. While subjecting Master’s student to similar conditions as researchers trying to publish in world-renowned academic journals is already a bit questionable, perhaps the worst part here is the “novelty” requirement. Usually, roughly 20% of your grade was allocatd to this component, which is effectively a vibe check for the professor on whether they think whatever you did was interesting. That this is relevant for proper research goes without saying, but expecting every single student on most of the department’s courses to produce novel research is simply unrealistic. Then again, in practice I feel like this is another component intentionally adding to the arbitrariness of the entire grading process.

As my final complaint before I talk about specific instances, I’d like to lament the lack of “inspection” like I had at home, though I’ve heard that other departments seem to do that. The CS department doesn’t do any of this. By default, you will receive the grade for your project work or exam, and that’s it. You have the option of mailing the department to give you the examiner’s comments (if they even left any) as well as the number of points you scored in each exercise, if applicable. The fact that these aren’t given out per default is… strange, and the transparency of the process is seriously lacking, leading to some unconfirmed suspicions that I’ll talk about later. This is especially annoying in combination with the research project exercises I’ve had to do: while I found the grading criteria questionable, I can’t deny the fact that the 3 research projects I did this way were by far the best practice and preparation for paper publication I’ve ever had. That makes it all the more frustrating that we receive no feedback at all on these, except for the final grade. That’s one sad, wasted opportunity.

Now, let’s get into the projects and exams. In my first term, I did Computational Learning Theory, Graph Representation Learning and Machine Learning. The former two were project-based, while the latter was evaluated via an exam. In second term, I did project work for Geometric Deep Learning and Uncertainty in Deep Learning, as well as an exam in Computer Vision.

Graph Representation Learning

This was a mixture between research project (40%) and proof exercises (60%). Can’t comment much to be honest, this seemed like a reasonable exam, though I only got a mediocre grade because my research project didn’t do anything the professor found interesting. I lost a bunch of points in the easiest parts of the proof exercises, which is completely inexplicible to me and I’ll never find out why because we don’t get any feedback =)

Computational Learning Theory

The CLT project was purely proof-based, and while I was a bit scared because the subject was hard, I did really well on their practice exercises after the very first sheet, so I was confident. Oh boy.

This was insanely hard. I spent 4 weeks working on this full-time, managed to finish only 60% of the exercises and got exactly the passing grade of 50%, meaning my submission kinda sucked but I tried hard enough that the examiner took pity on me. I still remember celebrating Christmas with my family on the 24th, and going straight back to my desk the moment everyone left, to do proofs for a couple more hours.

The set of exercises was quite strange in a way that I can’t quite explain, except that all the exercises felt completely different to everything we had done during the term. This feeling is somewhat supported by the fact that all my course mates did really well on the project: all of the… uh… 8 other people who did this got at least a merit, i.e. more than 65%, despite almost all of them struggling considerably more than me with the exercise sheets during term.

Strangely, half the exercises were nothing but pure maths exercises that sort of made sense in the context of CLT, but you didn’t need to have taken the course to do them. I’m pretty sure you would be able to pass without knowing any of its contents. That specifically doesn’t bother me, but it seems like a very weird design choice.

I’ll have to admit that I probably shot myself in the foot majorly by taking this course. It is my worst grade and likely cost me the overall merit, significantly damaged my mental health during my first few weeks in Oxford as well as keeping me very busy during first term, and, while being interesting, it turned out to be irrelevant for me.

Uncertaintly in Deep Learning

The UDL project consisted of reserach only, i.e. do some coding or proofs, then compile that into a 6-page report following the format and structure of an established journal. Specifically, we were given a paper to read, with the task to reproduce its results and extend it in an interesting way.

First of all, I was relieved when I saw the task, because I was worried it would rely on all of the things we ended up not learning during the disastrous teaching experience that was this course.

Then I saw an interesting statement in the introduction of the task… “This mini project should be completable within 1 or 2 days (excluding time spent on the writeup)”. Now this is either our lord and savior Yarin Gal showing some kindness to us students whom he failed to teach anything during the course, or it is a huge red flag, so fucking massive that it blocks out the sun and the sky as it covers the entire Earth, putting us into a dark, crimson depression. Of course it was the latter.

I think I spent 20 days working on this nonstop. The main difficulty turned out to be the “reproduce the results of the original paper”-part of the task, which I’d thought of as nothing but a warmup… boy, was I wrong. This task alone took me 2.5 weeks to complete, with the research component, you know, the actual main task, receiving the blessing of my attention for a mere three days. The ideas in the paper were a bit hard to decipher and turn into code to say the least, but even more so, I just didn’t get the same results as them. I became so desperate that I looked at the original paper’s code and started copying in random portions from there (as far as you can say that when adapting from Tensorflow to PyTorch) but even when the code was functionally equivalent to theirs almost line by line, my results were still several percentage points worse. I started seriously suspecting that the original authors had lied in their paper or something, but then I tested their code and it achieved the proclaimed results… At this point I gave up, accepted the difference and did some boring extension in what little time I’d had left.

I’d talked to several of my course mates, and every single one of them reported struggling with this reproduction exercise and spending several weeks on this, rather than the supposed “1 or 2 days”.

Geometric Deep Learning

In GDL, I probably did the most interesting research. Like UDL, it was a pure research project, though they gave us a dozen rough research directions with several more concrete ideas for each of them. This was amazing, simply because it meant my uncreative ass was less likely to be punished for this shortcoming by not having an idea the professor considered interesting.

This project was decent, but I had to run a lot of benchmarks for this. I think I had my laptop run a benchmark script to do expirements for about 80 hours straight, and that was just to get the final results, not even including any of the testing I had to do along the way. This is relevant because benchmark time means GPU time and Oxford didn’t provide any of that. More on that later though. This was the first time I had to come up with a way to visualize 4-dimensional data, but I think I did decently well, though coming up with the visualization made my head spin.

The final grading here was highly suspicious. About 10 or so people shared their grades, and it turned out that all of them fell into the 70 - 75% range. Now, no one was complaining because that’s pretty good, but it feels like the professor couldn’t be bothered to grade these and decided to just give everyone a score of 1D6+69 or something. Again, because there’s no inspection, we’ll never know =)

Machine Learning

This was one of my courses from first term, which ended at the start of December, but all exams were held towards the end of May/early June right in the midst of the thesis phase. This is why I am putting this after all the mini projects, which happend after first term (in December) and after second term (in March) respectively.

This was a really, really hard exam, though in a different way than the CLT mini project. Looking at the previous years' exams, I quickly realized I had completely underestimated the subject. In short, they were expecting you to have a deep understanding of every single concept covered in the course, regardless of whether they were discussed in detail and practiced in the homework sheets, or if they were just a small sidenote in the bottom quarter of a single slide. I think I spent over 3 weeks preparing for this exam, the most I’ve ever done for any exam in my life, and I still ended up scoring only 54% on it.

Again, several of the ratings seem weird because I lost points in some of the easiest exercises, but there’s no way to check how this could have happened. The exam was the easiest one of the past 5 or so years, but it still contained a couple ridiculous questions, like expecting you to write a complex proof that requires deep understanding of a technique that was mentioned on a single slide once. Questions like these were obviously made to prevent people from actually scoring close to full scores.

Computer Vision

This was, by German standards, the only “normal” exam on the course. It had a couple difficult or confusing questions but I would say that all exercises in this exam were realistically doable if you prepared. I think I had 3 days to prepare for this after writing ML, and that was easily enough since I’d paid attention during the course and it was well taught. I scored a raw 83% here, which was by far my best grade, but because you can’t have people getting good grades in a sane exam, everyone was downgraded. In my case, that meant going from 83% to a 75% grade. Did I mention that their grading scheme is insane?

Thesis Work

For the final block, let’s talk about my thesis work. You get the third term entirely to work on a thesis project (and exams) though this is still relatively short by most countries' standards, giving you a total of 4.5 months. You have to register your thesis project by the end of February, effectively requiring you to decide on the topic early in your second term. In my case, I found an interesting project in computer vision that was mostly hands-on, practical programming and model training. That was exactly what I wanted, rather than a more theoretical project.

After discovering that topic, I had to get in touch with the guy offering it, which ended up being one hell of a journey. I think I tried for about four weeks to reach him by mail, and I also tried to visit his office. Unfortunately, the CS department has a small building that is only accessible by going through the physic’s department’s buildings, which I couldn’t because my student card can’t unlock their doors… I only finally got a reply from him after complaining to the department’s administrative team. Perhaps this should have been a sign to look for a different project, but it was genuinely hard to find anything that caught my interest at all. Most other things were either highly theoretical or not about machine learning at all.

In any case, I sealed the deal after 4 weeks of trying, 1 week before the deadline. I then began working on this in third term, and two big problems became apparent. My project was about a new pre-training method for vision transformers, based on an existing pre-training method. The original method was trained on 64 GPUs in parallel for 4 days. I was worried we wouldn’t have the resources for that, but my supervisor assured me that wouldn’t be a concern, we could run it on the university cluster no problem. Once I finished the programming part and told them I was ready to run my experiments on the cluster, along with some initial estimates of the run time, they said, well that’s a bit much… Turns out we only get 8 GPUs on the cluster and (by design) our method is slower than the original, so we were looking at a training time of about 80 days for a single run. This was about 30 days before my submission deadline. We decided to shorten the training time from 800 to 50 epochs, so a single experiment would only take 5 days. We could run 5 experiments in parallel, and we could also train the original method in parallel, so we could compare this “early out” performance of our new method to the “early out” performance of the old method. If it’s not obvious, this sucks as far as useful results go, since it’s impossible to claim that one of the two methods is really better from the results of these experiments. What if my method performs better after 50, but it stops improving after 100 already?

The other problem was my supervisor. They were not that helpful and hard to get a hold of, if that wasn’t clear already. They added me to their research group’s Slack and the situation got better after that. They’d even proposed to hold a biweekly meeting before I could make the suggestion. But then they often ignored my messages, or took a week to reply, and perhaps worst of all, they didn’t appear to our biweekly scheduled meetings like, a third of the time. Several times I sat in the meeting for 20 minutes waiting for them, only to eventually get the message that they’re currently holding a lecture??? Dude are you telling me you didn’t know this 2 months ago when I scheduled this meeting?

All of this wasn’t such a huge problem for me because I could work almost entirely independently from them anyway. But once I’d finished my code and was ready to benchmark stuff, it got really bad. Students aren’t given access to the university cluster, only professors are. I was completely dependent on them to run my benchmarks for me. As you can imagine, this was a disaster. I was done with the coding part about a month before the deadline, there was nothing but benchmarking and writing left at this point. It took them 20 days to run a single experiment for me. And after that it took them another week to send me the results from this one experiment. I had literally nothing to show until three days before my thesis deadline!

The writing part itself was smooth, fortunately. It took me perhaps 4 weeks of writing in total, significantly longer than my Bachelor’s thesis, despite having almost the exact same length. I clocked in at about 16k words, so probably less than this blog article will have in the end (me proof-reading everything again: oh shit). In my humble opinion, this was a bad thesis, because I barely had any results to show, and couldn’t do any ablation studies or try to refine the method with better parameters. Still, I just got the grade (like, literally 4 hours ago) and I somehow scored 67%. Can’t complain about this honestly, except I would have gotten an overall merit, had I scored 68%… oof. But well, this piece of writing didn’t even deserve the 67% in my opinion, and I am so happy to be done. Supervisor-induced stress was a completely new flavor that I never needed in my life.

Miscellaneous Notes

Here goes a few more complaints or other notes that I couldn’t fit in anywhere else.

There are a lot of talks happening in the CS department, several every week, about various topics from our local researchers as well as visiting scholars. I think this is really cool, but I managed to go to only 2 of them in total. Still, it’s a great opportunity. It’s pretty cool if you read papers or something and you recognize the names of the authors because they’re in your department. I remember looking stuff up for one of my homeworks, ending up on an article on towardsdatascience.com, and recognizing the author… because he was my tutor. That’s pretty cool, and makes you feel like you’re at a place that’s at the forefront of research in the field.

The CS department has surprisingly bad equipment. There are only two lecture theaters in total, and none of them have sockets to charge your laptop. It would have been difficult for me to have three lectures in a row because my laptop’s battery couldn’t last that long. The building has a funny shape though: if you look at it on Maps in satellite view, you can see how it consists of separate buildings that were all kind of stuck together. People eventually tore down the walls between them apparently. As a result, the inside is a real maze, which is both impractical and kinda fun. Google Maps has the position of every single room on the 4th floor noted for some reason.

Perhaps one of the most annoying things was that we weren’t provided any hardware to train our models on. The department offers a number of ML-related courses, but you’re never told how to properly set up e.g. PyTorch, and instead to “just use Google Colab”. Google Colab is an online platform providing access to a Jupyter Notebook-like Python scripting shell with the option to connect to a GPU for free. This is problematic for several reasons. One is of course that we once again become dependant on a centralized software provider that has been shown in the past to be both quite expensive and not very trustworthy. Even more importantly, setting up a system to do ML with PyTorch is not exactly simple, since it adds CUDA version hell on top of the usual Python package management hell. I’ve had cases where I wanted to do homework locally rather than on Colab, and for some of them I just gave up, because it was seemingly impossible to get stuff to run on my laptop. The vast majority of my coursemates never went through this pain and thus probably don’t know how to set this up at all.

Worse though, while Colab provides all this for “free” there are usage limits, and those are entirely intransparent. At some point, Google will just say “no more GPU for you”. You can still train on CPU in that case, but, depending on the situation, this can lead to an increase in training time by 100x. I had a practical exercise to do at some point, which took about an hour to train locally on my laptop GPU. This is effectively impossible to do on CPU. I know of several people who trained this model once in Colab and were then locked out of training on GPUs for the next 2 weeks. This was a single, not all that relevant, practical exercise, so not too much of an issue. But remember how I mentioned that I had to run benchmarks on GPU for one of my graded projects for a total of 80 hours? As someone with a laptop GPU, I was the exception: most of my friends paid like 50€ in Colab GPU credits to be able to even do their graded project! Again, Oxford gets away without bothering with the problem at all, telling their students to “just use Colab” and offloading the costs to us. This topic was brought up several times by our course representative, but nothing changed until I finished my course.

The composition of nationalities in my cohort was quite interesting. We were a group of about 60, and out of those, to my surprise, the Germans were the second biggest group after the British, and even before the Americans and Chinese who shared the 3rd place at 6 people each. Funnily, 4 out of the 9 Germans there studied at RWTH Aachen for their undergrad. If the RWTH was a country, it would have been the 6th biggest group, right after Romania with 5 people.

The exams vs. mini project system leads to some questionable strategizing: you can choose freely which courses you take and whether a course is evaluated by exam or project work is known beforehand, so you can elect courses with an eye towards how many projects you want to take on each term and how many exams you want to write during thesis time. For the record, this transparency is good. What’s not good is that, depending on your course selection, you may be disadvantaged in different ways.

I’ve had several course mates who refused to choose courses that were evaluated through exams. Many people plain dislike exams in general, and most of them didn’t want to have exams overlap with their thesis work in third term. As a result, these people had to do at least 3 mini projects per term break. These term breaks were 6 weeks long, out of which the first 5 were used for project work. Projects were published on the last Friday of the term and the deadline was usually on the next Monday five weeks later, so we had about one week of free time between terms. Did I mention that almost every single person in my course who I’d talked to at the end of the thesis phase mentioned burnout in some capacity? This course was a marathon of 11 months with almost no break whatsoever.

Anyway, getting back to project work: these were, by department guidelines, supposed to take 3 or 4 days, and this was also usually mentioned in the task description of every project I’d seen. This is uni-speak for “about 2 weeks”, so you can imagine that it would get a bit difficult if you had to do more than 2 projects. And that was only an estimate: as I mentioned before, the CLT project took me 4 weeks, and the UDL project took 3. I only had two projects per term so it was fine, but I was still busy until the deadline in all cases. Some of my friends did 3 projects per term, and every single one of them had to ask for a 1 week extension in second term, as far as I remember. This means that those people worked through their entire 11 month degree without a single break, because this way they lost the one week of free time we would have had between project work and the start of the next term.

Conclusion

To conclude, I couldn’t say that the teaching experience at Oxford has been overly impressive. I’ve had good courses, I’ve had bad courses, I’ve had everything inbetween, and I think that’s normal for university. Overall, I would put it on a similar level as my experience in Aachen, but since I’ve only had 6 courses in total, your mileage may vary, depending on what you choose.

I can definitely say that the department was less well equipped than the RWTH was. I’m surprised a place as prestigious as Oxford that takes massive tuition fees is not in a better state. The department as a whole seems to suffer from several serious problems such as arbitrary and non-sensical grading schemes or an inability to provide their students with the hardware they need to do their homework.

That said, I did get in touch with a much more diverse group of students and very well established professors and other researchers whose names frequently pop up in the papers I read. This was pretty cool, but the actual studying experience was probably the weakest component of my year in Oxford.

Life in Oxford

For no particular reason, I’ll continue with my living situation, the vibe of the city Oxford in general, and use that as a spring board to badmouth the UK infrastructure for a bit.

Accommodation

As mentioned before, I got a (by my standards at least) extremely expensive student accommodation managed by my college. It was located in Summertown, a small suburb-ish area north of city centre, perhaps 25mins walk from college or 35mins from the city centre. It was a medium-sized accommodation, housing about 80 students divided into several “flats” that share an entrance, kitchen, toilet and shower.

I found out later that I was lucky to get into a flat with only 4 people: all other flats I knew of were shared by 6 to 8 people. I would say I was mostly lucky with my flatmates: a Danish PhD student, as well as an Indian and American guy doing a Master’s. They were pleasant neighbors, nice people, quiet and mostly clean. I later heard horror stories from a friend sharing a flat with… noisy, “noisy” and very unclean flatmates. In any case, we had a similar kitchen size as the bigger flats, though only a single shared toilet/shower room. The latter was difficult for me at times, since the bathroom was usually occupied for almost an hour on end on many mornings by three people taking showers and using the toilet in a row. If you have a bowels disease and strictly have to use the toilet within 30 minutes of waking up, this is… problematic. I quicky learned to adapt my sleep schedule and get up earlier every morning so I can use the toilet before the others get up. At least we had sinks in our rooms, which reduced the overal bathroom-occupancy. I don’t want to imagine the situation if we all had to use the bathroom to brush our teeth or wash our hands. I’m still a bit mad I paid four times as much here as for my undergrad accommodation and not even getting my own bathroom.

We agreed to share all our dishes, cutlery and other kitchen utensils, which helped with availability of stuff you needed and significantly cheapened my move to Oxford. That said, I barely cooked in that kitchen, so it didn’t make a massive difference.

Alright, let’s get into the fun stuff: my room. As mentioned, it was a room of about 20m² that came with a double bed, desk, chair, closet, drawer and sink. I didn’t have to get any furniture, which is nice, and the room was surprisingly spacious. This was bigger than my “apartment” in undergrad, even though that one came with a separate bathroom. At least this meant I had enough space to practice handstands excessively. In a similar way, the bed is weirdly big. Like, why would you provide students in small student accommodation with a double bed by default? The room was big enough for this anyway, but it still seems a bit wasteful. At least I was able to make use of this to save a visiting friend the hotel costs, but that’s about the only use I got out of it.

Perhaps the worst part, though, was not the room itself but rather its placement within the building: it was a groundfloor room with a window not to the inner yard, but towards the outside. The only thing separating my window from the sidewalk and then the road, was a small parking lot. That one was so close that, on many days, I couldn’t open my window without hitting a car, because someone parked immediately in front of it almost every single morning. Yes, the windows open to the outside only. The next issue was that my bed is placed right in front of the window, and there wasn’t really any other place to put it, despite the large room. This meant that when I was sleeping, my head was about 5 meters away from the road, or about half a meter from a car parking in front of my window.

Now, this was a residential street, but unfortunately it was the only one connecting two major roads, Banbury Road and Woodstock road, in about a 1km radius. In other words, a lot of cars were passing through there at every time of day, including a lot of freight trucks! This alone probably had me rediscover my hatred for cars and associated infrastructure. I didn’t sleep well during first term, because I’d always get woken up by massive trucks that began passing by my window at 6:00 every morning. This got better in the later terms, or maybe I got used to it, but I still lost out on sleep frequently due to the location. Most days I spent in some sort of half-sleeping state from 6:00 onwards.

There were a bunch of other fun events disturbing my sleep throughout the year. For one thing, I learned that apparently, car alarms in Germany are required by law to turn themselves off automatically after a few minutes. I’d never thought about this, but I learned that this was not the case in the UK, when one night I couldn’t fall asleep because a nearby car kept blaring its alarm for almost two hours straight. There was also a supermarket right around the corner from my accommodation, and they had their goods delivered… to the parking lot right next to my apartment. At least once a week, a truck arrived there at 7:00 in the morning, came down the road and stopped right in front of my window, then drove into their parking lot in reverse. Do you know what happens when UK trucks go into reverse? They have speakers loudly telling people to be careful because they’re going backwards. Ugh. Finally, I remember waking up early one morning because a mother and her child were in front of my room, and the only way I can explain the noise they were making, is that they kept hitting the asphalt of the parking lot with a metal shovel or something.

Finally, to get to the last downside of ground-floor, outward-facing rooms: I was completely visible to any pedestrians passing by, unless I closed the curtains completely. Which I did a lot of the time, but it was still a bit too sad to keep the sunlight out, and airing the room was more difficult when I kept the curtains closed. At least three times I remember jumping up in my desk chair because one of my friends living in the same accommodation knocked on the window right next to me, since they saw me on their way back from uni or grocery shopping. Now this was funny because they’re friends, but the idea that I’m this visible is kinda creepy. I wonder if college takes care to not give this room to women.

As I mentioned before, having only a single shower/toilet was a bit critical for me. Unfortunately, we got the accessible flat, meaning we had a huge connected shower and toilet area, compared to most other flats in the building which had smaller, but separate, shower and toilet rooms. That would have increased throughput significantly, but this way no one was able to go to the toilet as long as anyone was taking a shower. It also meant that most visits to the toilet were extremely humid in the mornings, since three people had taken their shower right before that. And since it was an accessible shower, it was not separated from the rest of the room by a wall or even a step, and for some reason the drain was in the center of the room, in the exact middle between the shower (left) and the toilet (right). This means that it was not only humid, but most of the floor except for right in front of the toilet was wet. There was a pullable string that would trigger an alarm, presumably for people with a disability to call help. I wasn’t the only one who pulled this accidentally once.

Next, the kitchen: this one was quite spacious for us and decently equipped, with a fridge, oven, stove, toaster and microwave provided, as well as plenty of storage space. There was only one problem: it had a window to the inner yard, but that one couldn’t be opened. One of my flatmates cooked there almost every day, and the humidity got so insane that we had water drops running down the walls a few times. My flatmate tried to air out the room towards the hallway, and thus triggered the fire alarm, forcing the entire 80 residents of the building to evacuate about 3 times during my stay there.

Let’s take a break with a couple positive things before I continue my rant: college was extremely responsive if I reported any breakages or similar. They were usually fixed without question after a single working day. This included things such as hot water in the shower not working, the toaster being broken or the faucets in the kitchen dripping. The internet situation was also pretty good: wifi connectivity was reliable, and we got incredibly fast wired speeds of about 30MB/s in download. That’s probably the fastest I’ve ever lived with. There was a common room at the top floor of the building, though almost no one knew about it. It was very big at a length of about 20m, had a TV, a collection of movies, books and even boardgames.

Alright, back to the negatives! What kinds of faucets do you think there are? I personally knew those with a sort of lever you could move up/down to adapt the pressure, and turn left/right to adapt the temperature. Those were the better ones imo, followed by those where you had two turnable wheels to change the pressure of the hot and cold water components respectively. You could effectively achieve any desired temperature inbetween by regulating the amount of hot and cold water in the mix that came out of the single faucet. I found out about a new kind in the UK that I didn’t know existed before: imagine the second kind, but instead of only separate handles for hot and cold water, you got separate faucets, so the sink has one hot and one cold faucet… what the hell? This is impractical, since the only way to adapt the temperature is to close the drain, then mix the appropriate amounts of hot and cold water in the sink. Otherwise, you can only choose between cold and boiling hot water. This was a strangely common type of sink, I saw it many times, not only in my room but also the kitchen and several other buildings. This seems siper outdated… except my student accommodation was built only 25 years ago! Though I guess I should consider myself lucky, since the shower had ajustable temperature. Imagine the horror. But hey, instead we had only a single dial that adjusts the temperature, so you always get maximum pressure. Not a problem for me, but, like, why? Look forward to the rest: Impractical or outdated infrastructure will come up a few more times throughout this post.

Since I began my studies in October, heating soon became a recurring topic. As it turns out, the buildings were poorly insulated, and at the same time college tried very hard to save on heating cost by turning off the heating in all student accommodations for most of the day. We were granted a total of 8h of heating per day: 3h in the morning, and 5h in the evening. Did I mention that the buildings are poorly insulated? During the winter days, I usually heated my room up to about 20°C during the morning hours, and roughly an hour after the heating turned off, we were already back to 16°C. The college sent us a warning at some point, that our flat is “too humid” and we should air more often for fear of mould. Did I mention that this was in peak winter? Perhaps the best part of this was the fact that my accommodation was still somewhat decently insulated, as one of the newer buildings. I had friends living in EPH, a student accommodation on the college main site, which had it much worse. That building was about a hundred years old if I remember correctly, and apparently people were freezing in there during winter. After weeks of complaints, college eventually relented and increased the heating hours to about 16 per day I believe, which is at least enough to get through the day. Some people in other colleges did have it worse though, I remember hearing about someone having a blackout that lasted almost a day. The last time I experienced a blackout was when I was 6 or something, this seems kinda wild to me.

All student accommodations had scouts which is apparently the Oxford-specific term for a cleaning lady. I am slightly conflicted about this: on the one hand, it seems a bit strange to be a poor student living in overpriced accommodation that is lacking in several ways, but then have what feels like a rich person’s luxury. I don’t exactly dislike my weekly cleaning routine, and I believe it is good to force young people to take care of their own household. On the other hand, this is definitely appreciated as it lets us students focus more on our studies. Additionally, I think it takes a lot of friction out of our shared living situation: people not cleaning up after themselves when they use the bathroom or the kitchen can become a serious issue. But if you have someone cleaning the common areas every day (except on weekends) that makes it easier to live together harmoniously. In addition to cleaning the common areas, they cleaned our rooms once a week, took out the regular trash, and provided fresh bedlinen.

Surroundings and the City

While my accommodation was a bit far from the city centre, Summertown was a very nice spot: I had less than 3mins walk to the nearest supermarket, and even three of them to choose between, as well as a bunch of other places such as restaurants, a shop for tech stuff, a pharmacy, bakery etc. It was well equipped with most things you needed, and at the same time it wasn’t nearly as crowded as the city centre. On a sidenote, self-checkouts seem to be very popular here, compared to Germany, which I liked a lot. When I returned, I realized that most supermarkets don’t have any self-checkouts here, and if they do, they sometimes remain completely unused even when there’s 12 people lining up at the only open cashier.

The only disadvantage of the location was the distance: about 2km to college and 3km to city centre. This is not massive, obviously, but it’s still a daily commute you’ll have to deal with somehow, and pretty far by Oxford standards. There were buslines nearby that took about 15mins to city centre, but few people wanted to pay for those every day. I have a friend who walked more than half an hour from the accommodation to college or their department every day. For myself, I got a used bicycle immediately on my second day in Oxford. This was only 70£ and, funnily, it was a better bike than the one I’d been using in Germany for the past 7 years. That is a low bar, to be fair, but still. The most notable difference was that it was incredibly fast. Riding to college took me only 7mins on average, and since this was so short, I could always pump hard. This was great, especially because it was my only real workout for most of the year.

What was not so great were the cars: on one of the two possible routes to city centre, there was no cycle path and cyclists had to ride on the road. It was a single lane in each directon, and I quickly learned to ride in the middle of the road at all times. If I didn’t do that, there was ALWAYS, every single time, some homicidal asshole that decided because the opposite lane was currently busy, they’d have to pass me by at a distance of 20cm. Whenever I forgot to ride in the middle of the road and instead kept to the left side, I was always, without exception, punished by one of those idiots and arrived in college both angry and scared. Did I mention that Oxford helped me rediscover my hatred for cars? It took me about half a year to realize that it was not worth it to take that road, and so I instead used a parallel one that was slightly longer but came without the risk and the stress.

I can’t comment much on the UK bus and train infrastructure: I used a bus within Oxford maybe five times in total, though it was nicely uncomplicated: it was a standard fare of about 1.50£ per ride, so you just slap your credit card on the scanner once when you boarded, and that was the entire ticketing process. I only used a train once in the UK, but I heard from friends a couple times that they’re always late. In other words, I felt right at home. Finally, one thing that deserves praise were the airline busses: Oxford has a bus company that runs busses at least every 30mins between Oxford and the two closest London airports, Heathrow and Gatwick. This means the trip from Oxford to the closest airport could be as little as 1.5h at a cost of 40£ total if you bought a return ticket. This was significantly faster and cheaper than if you’d go to London first and then to one of the airports: the trip to London was already 1h by train or longer by bus, and then perhaps another hour to get from the city to an airport. That second part could easily cost you as much as the trip to London did in the first place. Did I mention that flight subsidies are insane? There are flights between Hamburg and London for as little as 30€. This is barely more than the bus ride between Oxford and the airport!

To talk a bit more about the city in general, one thing becomes apparent rather quickly: buildings in Oxford are old. The people there are quite fond of their old buildings, and I understand that to some extent. They do have a certain charm, reminiscent of medieval castles or similar with their dark, imposing stone walls. But this comes with a bunch of problems: I’ve mentioned insulation before, and that’s only one of the issues of old buildings. They tended to be cold, poorly equipped in terms of sockets and faucets etc. or otherwise inconvenient. To me personally, they’ve lost most of their charm since I know how they represent the lacking state of infrastructure and housing in the UK. I’ve experienced first hand, and sometimes heard from friends, how these are not nice places to live in. At the same time, many of these old buildings close to where I lived were being sold for more than a million pounds. Prices in the UK, and Oxford especially, are insane.

Apart from that though, I really liked Oxford as a city. With a population of about 150k, it is a bit smaller than Aachen (~250k), feels more compact and significantly more lively. There is always something interesting going on in Oxford, usually all sorts of events from the university, departments, different colleges and societies etc. Oxford is a hotspot of activity and that was amazing. Its density means that you can quite easily reach most places by walking. In fact, you wouldn’t be much faster with a car or bus, since a big portion of the city centre’s core is a pure pedestrian zone, making it difficult to navigate with anything else. It’s a nice and convenient place to explore.

It it a very diverse and international city: you’ll find restaurants and stores of all sorts of nationalities here. I remember passing 3 Chinese supermarkets on a 50m stretch of road at some point, and there’s the market right next to the central bus station, which has a mixture of British, Korean, Indian, Chinese, Japanese, Jamaican, Brazilian etc. food, which is open 4 days a week. There is a Japanese restaurant which I’ve frequented with my Japanese-studying friends, that had authentic Japanese food (almost impossible to find in Hamburg) and a cheaper price than most of the British restaurants even. Moreover, the staff was almost entirely Japanese, so we ordered and sometimes chatted with them in Japanese - this was great. There are great Chinese restaurants as well.

The uni has its buildings spread throughout city centre and around it. There is a college or a department around every corner, though somewhat remarkably, there are almost no “shared” buildings in the city: if you are in X college and study Y, then you will rarely ever enter buildings that don’t belong to either X college or department of Y. Some of the libraries are perhaps the only exceptions here. I’m a CS student though, so I only visited the big old libraries like the Taylorean or the Radcliffe Camera on my very last day - I didn’t want to accidentally touch a book and burn myself after all.

Many colleges are open to visitors to come by and see. Most of them are quite charming from the inside, and there are significant differences between them. St. John’s for example owns like half of Oxford, or so it feels. In fact, even the grounds of my college initally belonged to them, and St. Anne’s hasn’t owned the land its buildings are standing on until somewhat recently. Magdalen is pretty big and they have a large lawn where deer live within the college grounds. Christchurch is one of the most well-known colleges, and is usually closed to visitors. You’d have to pay like 20£ to go in there.

In case that didn’t make it obvious yet, Oxford is quite touristic. I wasn’t expecting this at all to be honest. While the city is certainly nice and has a couple pretty colleges and old buildings to look at, that doesn’t seem very special to me. Yet the city centre is sort of overrun with tourists for most of the year. Especially in summer you go through periods of hearing mostly Indian, then French, then Japanese and then Chinese from the tourist groups of gradually shifting nationalities. The Ashmolean library is open to both students and tourists, but at a hefty price for the latter.

During my time in Oxford, I visited 13 out of 38 colleges if I counted correctly. Various societies have their activities spread out across different colleges and some events are open to everyone, I visited friends to cook or eat together in their college, or went “as a tourist” to look at other colleges. That said, many of them are still quite similar, and I don’t think you gain much out of visiting every single college.

In fact, when a friend from Germany visited me in Oxford for a few days, we finished the “sightseeing” part in about 6h on a single day. Of course that depends on your preferences and there is certainly more to see in theory, but as a visitor, it’s probably enough to walk through 2 or 3 colleges and city centre for a bit, because there’s not much new stuff beyond that. You’re not seeing the famous libraries from the inside, because they’re either completely closed to visitors, or come with a 20£ entrance fee.

On a sidenote, every single college has a main gate guarded by their porters 24/7. They’re essentially guards keeping out unwelcome visitors. This has always felt a bit strange to me, because it makes colleges feel like walled gardens, though I guess this is quite fitting.

UK Vibes and other Notes

Most communication among students in Oxford happens through WhatsApp, and a sizable portion of it through Facebook. Almost all my friends are on Instagram, effectively turning this app I used once every couple months to brag about my mediocre bouldering skills into something I check every couple days to see what my friends are up to. As a privacy-conscious software person, I shouldn’t be using Instagram at all, but here we are. I’d finally uninstalled WhatsApp a month or two before coming to Oxford, after spending years trying to get my friends and family onto Signal or other platforms. And only days after getting there, I installed it again, since I would have been excluded from a lot of the social stuff going on. There are events and societies organized completely through Facebook, but fortunately there wasn’t anything important on there. I usually didn’t bother in those cases. I sometimes find it hard to believe that people still use Facebook, because it has been so obviously absent among people in my social circle for most of my life. The Americans mentioned getting WhatsApp specifically for Oxford, because they don’t use that at home. I was rather surprised to learn that they mostly use the Apple messenger, implying that they all use iPhones over there. I had no idea this was the case, and it’s genuinely horrifying to me.

There’s one thing I never quite got over during my time in Oxford: my fight-or-flight reflex kicking in when random staff shoots a “how are you” my way. This is mostly cultural difference: “how are you” can be a greeting in the UK, requiring either a stock “I’m fine” answer, or no answer at all even. But I am German, and when one of us asks you “How are you?” they are prepared to hear your entire lifestory. Then again, that might just be me. When I ask someone how they are, I expect an honest answer, or at least want to give them a sincere opportunity to tell me about anything that’s on their mind. I’m never going to ask this question to someone who I don’t care about. Something about asking it and neither expecting nor being willing to listen to an honest answer seems depressing to me. This feels like it puts pressure on the other person to smile and say “All good!” when they might already be having a horrible time. I’ve refused to ask this “casually” for all my life and at least the people around me are sort of similar. As a result I’m used to answering this question seriously myself, so whenever I get asked this question by staff, it’s like I have to forcefully interrupt my regular train of thought before I accidentally answer the question seriously. While I’m all for taking it seriously, I’m not going to bother a random cashier with my worries.

To give you my bottom line for the UK experience: I’m not impressed. Everything feels too expensive, cost of living is super high for no good reason, infrastructure is in a bad state and most buildings are not great places to live in. It feels in many ways like a toned-down version of the US (from what little I know about the US) and, from my personal experience, a straight downgrade from Germany in most regards. This was echoed by many of my friends and acquaintances there: only a handful were planning to stay in the UK, mostly because going from student to work visa was relatively easy, and they were already speaking the language. Most said that they would prefer going somewhere else after they finish, and several people mentioned that they found the UK a pretty sad place. I at least liked the city itself, as an incredibly diverse hotspot of social acitivites.

Food, FOOD and food

I want to talk about food some more, since there are interesting comparisons to be made.

Overall, food in the UK is not very good. As someone who barely uses the internet (heh) even I’ve heard of this before coming to the UK, but it is still interesting to see in what way UK food is bad. But for context, let me first tell you a little about my eating habits during my stay. I usually had one meal consisting of overnight oats with apple and banana, which I made myself several times a week. Then there was the college dining hall, which I ate in once almost every day, sometimes twice. I sometimes bought sandwiches from a supermarkt, or ate a simple slice of bread with cheese. I didn’t eat out during my first term, but later I started going to the market in city centre or to a Japanese restaurant with my Japanese study friends once a week, sometimes cooked with friends or went out for food with other friends, got food at college events or similar.

The main way in which food in the UK tends to suck, is the lack of spices. This is especially true for dining hall food, which often seemed to lack even the tiniest pinch of salt. This was strange, because the food often sounded good on the menu, and frequently even looked good in reality, but it tended to taste like nothing. The dining hall experience could be significantly improved by adding just a tiny bit of salt. I don’t understand why this was the case, because they provided us with salt dispensers etc. for free on the tables. But many foods don’t taste as good if you only add it at the end.

One strange occurrence greatly exemplified this: there were lunch and dinner menus in the dining hall that changed on a weekly basis, as well as a breakfast “buffet” that was pretty much the same every day. One of the options there was scrambled egg, and I love scrambled egg, so on my plate it went every time I came there for breakfast. It was probably the worst scrambled egg I’d had in my life, because… well, unsalted scrambled egg is really just a mass of nothingness. It was so bad, I found myself wondering how they made eggs have this little taste, like, I’m pretty sure even just an egg with nothing else should have more taste than this… But on one unusual 29th February morning, they blessed us with decently salted scrambled eggs. You can’t imagine my delight when I tasted those eggs. I had no expectations in that moment, and it’s been months since I’d had decent eggs. It was so good I immediately messaged my friends to tell them about it. One of them came by and was able to share in the joy with me. But this must have been some sort of accident, because it never happened again afterwards. Hah…

The dining hall food was expensive by the way. For 2.90£ you got the main dish, for which you could choose between a meaty/fishy, vegetarian and vegan option (which was nice) though that dish alone was only very little. Think, a single thin, hand-sized schnitzel or similar. You could usually choose between 3 sides, each of which would cost 0.80£ extra. What one should consider a full meal would be a main dish with two sides, so 4.50£, or about 5.80€. Compare that to my undergrad where I paid between 2.10€ (vegetarian) and 2.50€ (meat option) for a full meal. Those meals usually filled you up and you could get a free refill of the sides even! This couldn’t be said about the meals in Oxford for the most part. There was one exception though: fish and chips Friday.

Every Friday was fish & chips Friday, and the fish was massive. On those days I usually picked only one side, and was still full afterwards. That was nice, but at the same time… this was fried fish in oily batter. It was probably the most unhealthy dish on the menu. Most of my friends and me as well, got tired of this meal after first term and never touched the fish option again on Fridays, which is a shame. I still remember a friend telling me how he’d smelled the fried fish from the opposite side of the college’s back entrance, which itself was located on the opposite side of the college as the dining hall. TL;DR it was a bit much.

The one thing the British (or at least the dining hall) seem to do quite well, is desserts. We frequently got desserts for like 1.50£ that tasted great and looked quite nice. Here’s a couple examples:

One last story about the dining hall, because I found it particularly funny: on a certain day in second term, I saw that there was “Curry Wurst” on the dining hall menu. Not “Curry sausage”, but “Wurst”, the German word for sausage, suggesting that we are indeed getting something resembling the German Currywurst. I wasn’t particularly excited about that, since it’s more of a fastfood and not exactly one of my favorite dishes anyway, so I was going to get the vegetarian option most likely. I lined up in the queue, and was immediately approached by this German girl whom I’d talked to perhaps 3 times before. She gave me a 5 minute rant in German how she’d seen the “Curry Wurst” at the front of the queue, and found it a blasphemy that had shooketh her to her culinary core. I honestly didn’t think much of this, since I didn’t have any expectations and wasn’t going to eat it anyway. Then I reached the front of the queue and got to see their “Curry Wurst”. They gave out sets of three Nürnberger mini-sausages topped with what looks like an Indian style yellow curry sauce… WHAT THE ACTUAL F-

Next, let me tell you about sandwiches. In Germany, we barely eat “those” kinds of sandwiches, we usually have proper baked bread stuffed with all sorts of fillings, such as cheese, ham, fish, vegetables, etc. sold at bakeries at any street corner in the country. The equivalent in the UK, are sandwiches, i.e. two slices of “toast bread” enclosing their filling. These are not all that bad, but, as a German, I do miss “proper” bread. I honestly don’t know all that much about bread, so I can’t be more specific, but toast bread feels a bit sad to me. In any case, they liked to prented their bread is a bit healthier by using “malted” bread which, as far as I can tell, makes the bread slightly darker so it looks closer to grey bread, and nothing else. On a side note, I saw something literally titled “German Rye Bread” in a supermarket, and that made me very happy. The only good bread I encountered during my time there.

In any case, on to the fillings: there was a wide variety, such as tuna & mayonaise, tomato & salad, curry chicken, ham, salmon & cream cheese, bacon & egg etc. But again, the lack of taste struck pretty hard here at times. Think of tuna and mayonaise, this is one of the cheapest flavors around at a price of perhaps 2£. It was profoundly sad, as you once again realize that the British can even make TUNA and MAYONAISE, the two strongest tasting foods from my list above, taste like nothing. That is not to say there weren’t decent sandwiches: the famous BLT (bacon, lettuce, tomato) for example tastes surprisingly good. But often times, it is impossible to tell from the ingredients or looks of a sandwich whether it’ll be any good. Besides, sandwiches are a bit expensive… like anything in the UK, but still. They start at 2£ and go up to like 5£. The latter is what I pay in Germany for a big bread from the bakery that is generously stuffed and has a lot of taste. That’s not always the case in the UK. Moreover, I checked for sandwiches in a German supermarket recently, and you get almost the same range of fillings as in the UK, except every single one of them costs only 2€, i.e. 1.70£.

The UK has a wide range of potato chips on offer, but for some reason it was almost impossible to find my good ol' paprika chips. For context, those are the “default” flavor in Germany: if you think of potato chips, you most likely think of paprika chips. I assumed it would be the same everywhere, but that’s apparently not the case. I tried a bunch of different flavors during my first months in the UK, but none of them did it for me. Then I learned by chance that the tiny convenience store right opposite of my college, of all places, had paprika flavored chips, so I went there frequently to get my fix. At some point I ordered a 1.5kg pack of them online. Life was good. Oh, but they’re incredibly expensive… about 3£ for a 120g pack, compared to ~2.50€ for 150g in Germany. Wow…

The selection of stuff in supermarkets was mostly good and I found what I needed. I did bake a few times but I never found the “vanilla sugar” that my German recipes required… now that I think about it, that does sound like an oddly specific ingredient, so that was interesting to notice. What I did find was shortbread, which I’d always liked. I couldn’t say that it tasted better in the UK than the one I’d bought in Germany, but it was significantly cheaper. You could get a 100g pack for a single pound, compared to 2 or 3€ for a pack in Germany. I also really liked Scotch Eggs, which are hard-boiled eggs wrapped in minced meat and breading. I found them way too late, but by the end of my stay I was eating several a week, and it is one of the few typical UK things I miss a little.

I will talk about this in more detail in the respective section, but one important part of college life is the “formal dinner”, or “formal” for short. Formals are fancier dining hall dinners with 3 courses, which have to be booked beforehand and are, of course, more expensive. Those formals were likely the only opportunity where I tasted decent British cooked food. Moreover, St. Anne’s formals are pretty cheap, at 16 - 18£. That’s one of the few times I genuinely couldn’t complain about prices in the UK. They put quite a lot of effort into arranging the food nicely. Here’s pictures from one of the formals I went to:

Finally, perhaps because the British are aware of their lack of food culture, there is an abundance of authentic foreign restaurants and other places to eat. I love Chinese food for example, but in Hamburg it is difficult to find an authentic Chinese place. Instead, there is an abundance of “Chinese” places, usually run by Vietnamese people, selling the typical sad, fatty fried noodle boxes. You barely see those in Oxford, and more importantly, there are at least two great, authentic Chinese restaurants in Oxford, a small town of only 150k residents. It doesn’t stop there: there is the amazing Japanese restaurant Edamame I mentioned before, which offers a small list of classical Japanese dishes, such as Ramen, Tonkatsu, Karaage, Yakisoba, etc. It’s already difficult finding any Japanese food in Germany at all - I’ll have to take 45mins of train rides to get to the one good Ramen restaurant I know of, at the other end of the city. But here, it was perhaps 15mins of cycling from my accommodation, and their food was even among the cheaper options for eating out in Oxford: 12£ for a good Tonkatsu complete with sides? Hell yea!

There is a wide variety of international food offerings in Oxford, so I’m going to bring up a couple more to give you an impression: there is a Portuguese Nata place where they sell pasteis de nata, as well as donuts with the classical nata sweet egg filling, nata-flavored ice cream etc. There are several Nando’s, an “Afro-Portuguese chicken chain” by their own description, which my brother liked so much he went there at least three times when he visited me. There are about 20 bubble tea (or rather, naicha/tapioka, as I would call them in Chinese/Japanese) places in Oxford, which I frequented with a certain group of friends. And then there’s the market in city center which has classical Japanese food, amazing bibimbap, several Chinese stalls offering different local cuisines, Brazilian and Indian food, Korean-style hotdogs, and several more that I didn’t get the chance to try, almost every day of the week. All in all, there were a lot of options for food in Oxford, though rarely cheap.

College and MCR Life

Alright, let’s finally talk about college life in more detail. I mentioned before that it’s a bit difficult to explain the concept of a college, because you only really get it through having lived there. The most important part for me, is that college tries to provide you with a sort of social core of people to hang out with, events to enjoy and connect with new people etc., and to a lesser extent, it is responsible for a lot of the administrative stuff and other tasks related to your student life.

I believe the dining hall I mentioned before is a good starting point: most departments (certainly not mine) don’t have dining halls, so the best option for uni-subsidized food is the dining hall of your college. This effectively makes my “local dining hall community” much more constrained than those during my undergrad, which probably served a couple thousand students every day, of which I knew less than 1%. In college however, I almost never ate alone: there were always faces I recognized in the dining hall, and for most meals I sat down with friends and had a good chat over lunch, or sometimes, dinner. I would even go as far as saying that the dining hall was the most important component for my social life in Oxford. I’ll talk about this in more detail in a later section, but to me, one of the most important parts of forging genuine, deep connections to people, is just talking to them a lot. The dining hall provided the perfect place for that. I would say my closest friends are likely the ones I met in the dining hall several times a week. Most of my meals there lasted an hour, and sometimes even 2 hours until we were kicked out of hall because they were closing, since I enjoyed chatting that much.

The next most important part is probably the MCR and all the events and activities associated with it. As mentioned before, the MCR is the group of postgraduate students at a specific college, to be distinguished from the JCR, which is the group of undergraduate (i.e. Bachelor’s) students. The MCR can be seen as its own organization, as we have our own elected MCR committee with its own consitution, the MCR has its own common room (also called MCR btw) and we essentially organize our own brand of events that is targeted only at MCR members. Our MCR committee consists of several roles such as president, secretary and treasurer and receives a couple thousand pounds every academic year, that we can use to organize activities and events etc. We have social secretaries and welfare representatives who are sort of the “core event organizers”, as well as several minorities' representatives, such as disabilities' rep, women’s rep etc. Social secs and welfare reps are essentially free to organize what they want (and so is anyone not in an elected capacity, to be fair) and in our case they managed a wide range of events.

Perhaps the most noticeable ones were the so-called bops, essentially big parties held in the MCR, and one time even in the college dining hall. Think loud music, dancing, lots of alcohol, sometimes with costumes etc. But we also had boardgame nights, movie nights, an UEFA viewing for Spain vs. England, an Alpaca petting event, pebble painting, yoga sessions, shut up and write sessions (get work done, while MCR provides you with snacks), karaoke nights, welfare tea, pub quizzes, Chinese new year’s dumpling making, Palentine’s, a bonfire/barbeque/live band event on a meadow outside the city, and probably many more that I couldn’t remember just now. We had so many amazing and memorable events, some of them very creative, and all of them organized by fellow MCR members who put a lot of work into it to make them fun, memorable occasions.

It is worth mentioning that there was a relatively clear split between the kinds of events organized by social secs and by welfare reps. The former were typically the “big party” sorts of events, while the latter were directed more at smaller groups and quieter events. I love the fact that we had both of these facets represented well, though if you know me, you’re likely not surprised that welfare events are more my style.

As described earlier, the MCR committee consists of elected officials, and while I didn’t feel up for such a position when I was only there for a one year course, I was interested in committee work in general. I had been part of my student dorm’s network group in undergrad: we maintained the entire ~400 people dormitory’s wired and wireless network, expanding it by putting up new wireless access points etc. as well as managing several other appliances we provided to other dormitory groups. I also became one of the two dorm chairmen for about a year. In short, I’d done… not sure what to call it, but this sort of thing before, and I was still interested, so at the very least, I showed up to every one of the committee’s monthly meetings. These were open to all MCR members, but there weren’t any non-committee members who showed up consistently except for me. This way, I still ended up engaging with the college’s political life, which I found interesting, and helped out with the preparations for some of our events, as well as hosting two karaoke events during my time there. At some point towards the end of the yaer, the president proposed to make me an official member, essentially giving me one of the unfilled positions so I could get committee member benefits, but I didn’t have to take on any responsibilities.

At this point I want to say something about the committee’s vibes, and I believe they reflect the overall St. Anne’s MCR vibe really well: people were incredibly nice. All of the committee members genuinely cared about the committee work they did and their fellow students, they tried hard to organize fun events that were inclusive and accessible to anyone, including people who were more shy or not the type for big parties. I’ve had a good life so far and can’t complain about my social circles, but I’ve never been in an environment before that felt so genuinely caring at all times. This is difficult to convey in words, so I’ll give one example that I believe illustrated this well: we had a meeting once where we had a rather uncomfortable discussion about a political topic, brought up by several non-committee members who showed up as a group to one of the monthly meetings. I had an appointment, so I had to leave the meeting early after about an hour, in the middle of the discussion. Shortly after, I got a message from one of our welfare reps, asking me if I was alright, and that they’d be there to talk if I needed to. And perhaps the most notable thing about this interaction was not just how they reached out, but how natural it felt. I can imagine something like this happening in a different context as well, but I think in most environments it would still feel just a bit strange. But that was not the case here, it was heartwarming and very in-character.

There were 18 committee members in total, and over time, I got the opportunity to have a chat with almost every single one of them, and would call about half of them my friends - that made committee work more fun to engage with, because most of the time it felt like helping out a friend. Similarly, I knew I’d see some of my friends at almost every event I went to. I met a lot more wonderful people in our MCR (that were not part of the committee) and I really loved our MCR community. It made me feel very welcome. Because I knew so many people here, it made college feel very alive to me. There were always people around that I knew and I constantly ran into friends around college, giving it a very homely atmosphere. This can also be said, to a lesser extent, about the city as a whole: I knew so many people here, that on almost every walk through the city centre, I met someone I knew.

An important part of my college life was the MCR (the room) and the study room. They were located in EPH, one of the postgrad accommodations, but open to every member of the MCR. The MCR itself was a large room, perhaps 50m², containing several couches, a big smart TV, a massive bluetooth speaker (like, >1m tall), two guitars, a big teddy bear, a table tennis plate, a shelf full of books and a box with boardgames, etc. In short, there was a lot on offer to engage with, and it was frequently used for MCR events, or just by me and my friends if we felt like doing a movie night or similar. I came by often during the first two terms for a break between lectures, especially in second term where I practiced guitar there almost every day. It was a great hangout room, which is especially important in Oxford where almost no one invites people over to their apartment since we’re all in tiny student accommodation or in a flat share.

The study room was similarly nice. It wasn’t super well equipped, but it had a comfy atmosphere, seating for about 10 people and was mostly quiet. It was a great place to do my homework and related work, especially when I had breaks between lectures. Between first and second term I shifted from doing most of my work at home to doing most of it in the study room. On most days, I spent only about 12h at home, and the remaining 12h in the department or in college. Thus, I became one of the study room regulars. It is comforting, in a way, to know you’ll find familiar faces whenever you go there. And while it was, of course, a place to study and be quiet, when it was only you and one or two friends in there, that often made for memorable conversations about anything and everything as a welcome distraction from studying. It is another part of the college that feels like home, because I spent so many hours in there, and I had a very strong connection to most of the other people who did the same. Among these are at least two people who I’ve barely ever seen outside the study room, but we just know each other due to spending possibly hundreds of hours together this way. I still remember one of them messaging me that I forgot my bicycle helmet in there. They weren’t there when I left, so I guess they must have recognized it as mine from seeing it so often. We also had a decently equipped kitchen in the building, which is quite convenient, and we used it for cooking and events several times.

There was one event in particular which I’d like to mention: Palentine’s, which comes from “Pal” and “Valentine’s”. The idea was that you could write a message (anonymous or not) to anyone in college, and committee members would prepare them together with a bag of sweets and deliver them to that person’s pidge (essentially a small mailbox) on Valentine’s. This could be romantic of course, but absolutely didn’t have to be, thus the “Pal” part. To be honest, I’m probably not the type of person who would usually participate in such an event, but this time I had a lot of people I appreciated in different ways whom I wanted to express this towards, so when this was announced I was immediately hyped and wrote a ton of them. It was one of those events where I helped out in the committee, in this case to prepare the bags of sweets etc., so I know that about 70 of those were written in total. I think it was a wonderful event idea, and I know it made some people, including me, very happy. I had the messages I received pinned to a board in my room until the day I left Oxford.

JCR vs. MCR

I found it quite interesting to observe how clearly JCR and MCR are distinguished and separated within college. For one thing, as mentioned before, they have completely separate committees and separate rooms. As in, JCR members are not allowed to enter the MCR (room) usually, and the other way around. They organize their own, usually separate, events although collaborations exist of course. This clearly separates most of our social spaces, except for the dining hall. And while St. Anne’s is an incredibly social space, you rarely sit down and start a chat with a random group of strangers in the dining hall.

The undergrads had almost completely separate accommodation as well. While MCR members had two small (~13 rooms each) accommodations on the main site as well as RSH in Summertown (~80 rooms) where I lived, the undergrads had at least four big buildings within college grounds, perhaps 100m from the dining hall. Moreover, their rooms were significantly nicer than ours. This was aptly illustrated by the fact that the fourth undergrad accommodation was currently being modernized, so college decided that half the flats at RSH would be allocated to undergrads instead. I later talked to an undergrad and thus found out that RSH, which is by far the most modern accommodation offered to MCR members by college, was easily the last choice for the undergrads, since their accommodations were so much better. At some point during my first weeks, where I didn’t yet know that these buildings were undergrad accommodations, I happened to come across two girls who were trying to lift large cardboard boxes out of a wheelbarrow in front of one of these accommodations. I assumed they were moving around materials for one of the college’s societies or something, and they were obviously struggling, so I offered them my help which they gladly accepted. I then carried several of these boxes into a room on the 3rd floor of said building, and thus learned that this was undergrad accommodation, and that I was helping one of the girls move in. The most interesting takeaway here was that the room was super nice, roughly what you’d expect from a two-star hotel.

Moreover, they told me that they had to move out in the summer of every year, put all their stuff into storage somewhere, and then move into a new room at the start of the next academic year. And apparently the four buildings were designated for up to four-year undergrad courses, where they got “the next nicer building” with every year they spent there. This is a bit of a strange concept to me. Moreover, apparently college forces out all their students through the summer break, which lasts 3.5 months in Oxford, so they can rent out their student accommodation to conference visitors etc. and make big cash.

Finally, undergrads have a surprisingly different vibe, and you could usually tell within seconds when you were talking to an undergrad, or often even just by looking at them. I’ve been told that you can distinguish between JCR and MCR parties by the presence of an ambulance. Undergrads would always go overboard, especially with the drinking aspect, or so I’ve been told. Many people seem to think that the clear JCR/MCR separation is for the better, and I’m somewhat inclined to agree.

This does raise an interesting question though: why are undergrads treated so differently? Many people I’ve talked to mentioned that most postgraduate students are 1-year master’s students, followed by 2-year master’s students and then PhDs, compared to undergrads who are staying on for 3 years and sometimes even 4 years. They argue that undergrads are treated favorably since they pay tuition fees about twice as high as the average MCR member. But I don’t think this argument holds: PhD students, who usually stay on for 4 years, comparable to undergrads, are not treated any differently than other MCR members. Moreover, international students pay more than twice as much in tuition fees. For my course, at the time of writing it is 17,900£ for UK residents compared to 41,250£ for international students. For the undergraduate computer science course, the disparity is even worse at 9,250£ vs. 59,260£ per year. The reason this matters is, that, according to the university’s Student Statistics, about 23% of undergrads are internationals, compared to 67% of postgrads. In other words, I can’t help but feel like they care more about UK students, in addition to significantly milking their international students for insane amounts of money. I didn’t know the tuition fee disparity was this horrible until I looked it up just now to make my point.

St. Anne’s Vibes

Next, let me give you some background on St. Anne’s: it is one of the former women’s colleges and one of the youngest colleges in Oxford, having become an official college only in 1952. It is the fifth biggest college by number of students, standing at 860. As I mentioned before, I didn’t know what to expect from colleges, as a concept, so I didn’t put down any preferences. I now believe I got very lucky with St. Anne’s. I’ve already mentioned how comfortable I felt there and how many wonderful people I met, but that tiny bit of history I brought up also has one important implication: St. Anne’s is far from conservative.

As someone who’s not from around there, I (and many others) had this vague fear that Oxford would be full of entitled, arrogant, right-wing rich-kids. This fear is… probably… not unfounded: from the stories I’ve heard during my stay these people absolutely do exist. There’s a reason the Oxford Student Union invited Ben Shapiro to talk about climate change, or that you hear about Christchurch students going to other colleges, intentionally breaking things and then dropping a thousand bucks on the spot to pay for the damages. I don’t even know if this story is true, but the fact that everyone believes it could be true is telling enough. But fortunately, at St. Anne’s you don’t see any of that. In fact, we are both known as one of the poorest colleges (bottom 5 colleges, statistically speaking) and as “the friendly college”. I’ll have to say, I never had any downright negative experiences with other colleges, but the “non-conservatism” of St. Anne’s extends beyond vibes.

For one thing, ours is one of the few colleges without a chapel. Can you imagine? Not that I have anything against churches, but the idea of having one on our college grounds still seems strange to me. Probably for historical reasons, we have one of the highest female-to-male ratios at our college, and our MCR committee in particular was like 70% women I believe. Perhaps the most noticeable difference though, was that we rarely had to wear subfusc. Subfusc is a dress code encompassing something like a regular suit, plus the gown. The gown is sort of like a robe, very traditional Oxford clothing that is only sold by three stores in Oxford. Subfusc has to be worn for matriculation and graduation, but at most colleges, also for formals.

At St. Anne’s, you fortunately don’t have to wear subfusc for formals, and not even a suit technically. The dress code was “smart”, essentially meaning “just wear something nice”. If you know me, you’ll know I don’t feel all that comfortable wearing fancy clothes, so I was quite relieved. St. Anne’s was, seemingly, less attached to formals than many other colleges. We had four formals per term, i.e. one every two weeks, compared to other colleges which usually had one per week or even more, from what I’ve heard. They were up to twice as expensive as ours, without a noticeable change in food quality.

Mostly due to dress code, formals are the kind of event that doesn’t quite fit my taste, but it was a good opportunity to get a decent meal at a good price. I frequently coordinated with my friends to ensure that we sit together, often turning it into a very social evening. Since you could see a list of all the names and their respective seats for those who’d already booked when you were visiting the booking page, this served as a great opportunity to catch up with people I hadn’t seen for a while by choosing a seat next to them, or the other way around, for others to seek me out. All in all, formals contributed nicely to the social environment in college. They are also one of those things that feel “typically Oxford”, carrying a sort of special vibe.

This is even more so the case at other colleges. I had the chance to experience this, since I made a sort of exchange with a group of friends: each of us invited the other 3 for a formal at their own college. It was, again, a great opportunity to socialize, to check out the food and soak in the vibes of different colleges. Their dining halls were typically much older than ours, and thus had a certain atmosphere to them. At one of these exchange formals, their college choir even did a (short) performance for us. One of my friends took it on himself to give us a tour around his college while telling us about their history. This was a great way to explore a different college and I’m thankful for the opportunity.

As another typical Oxford thing that I didn’t know where to put in this article, there’s the “Fresher’s flu”, which is a chain of colds going around at the start of every academic year in Oxford: as new and old students come to Oxford from all over the world, they bring with them viruses from their home countries. Due to the sheer variety as well as heavy socializing and many events at the start of the term, several colds spread throughout the entire student population of Oxford at the start of every year. My personal experience with this was that I essentially “had a cold” for the entire first term. In reality, it was at least 3 colds that passed the baton during the first 8-week term. Almost everyone goes through this experience, and it contributes to making your first term at Oxford a bit more miserable.

It’s worth talking about matriculation and graduation quickly. Both are massive ceremonies hosted by the university, in the Sheldonian Theatre. It is an old building of over 300 years, and has a certain charm. They really like their ceremonies in Oxford, so you’ll need to go through matriculation there under a strict dress code, listen to the vice chancellor and some latin phrases while an old man waves a big golden scepter around etc. We had to be very careful to follow the dress code, and I had one friend who was sent away at the entrance because their shoes were dark blue, instead of black. They graduated a few weeks later with like five other people if I remember correctly. While my own graduation is still about 2 weeks away as I’m typing these lines, I’ve seen the ceremony already when I accompanied a friend at theirs. Again, it is a very Oxford thing, but it drags on for quite a while and is a bit boring if you ask me.

Getting Milked

I mentioned earlier that it often felt like college is milking us for money. Here go a few more examples of this:

College organized a photographer for our matriculation day. They took a photo of the entire group (about 300 students) and let us take personal photos, one person at a time. If you wanted even just the group photo, you had to buy it. The cheapest printout started at 27£, with the largest framed option being over 200£. But the worst thing is, there was no digital version. I only wanted a digital version to share with my family. Now I don’t even have any photo at all, because I didn’t want to pay that much for it.

Remember when I mentioned the rent and tuition fees and how they increased over last year’s? I checked the tuition fees again, and it turns out they increased by a third for my course in the meantime - from 31,000£ to 41,000£! I’ve heard rumors that rent increased by another 15 or 20% as well. This is insane, I thought my year was unlucky but if this keeps going, I feel like people will stop studying there altogether. And for those that still do, college accommodation will be useless because it is both worse quality and significantly more expensive already at this point than private accommodation.

Did I mention that you can get a recording of your graduation ceremony… for cash of course! The idea of selling things for money that could usually be digitally reproduced at no cost whatsoever often baffles me, and this is one of those cases. The ceremony’s recording costs 20£, which is not even that much (I was expecting worse when I found out) but still… why?

St. Anne’s members received an email in early December, containing this:

There’s so much wrong with this, I don’t even know where to start. The pitch itself is super weird: who is looking for Christmas gifts, then gets this mail and goes “oh yea, sure, this makes for a great gift”?? Imagine gifting someone their name engraved into a paving stone at your college, just… what? And then the price for this, how does this come to 2,500£? And if I understand the part at the very bottom correctly, then only 500£ of this is the actual donation, while engraving the pavement stone costs 2,000£? What the hell are they doing?

To take a tiny step back, I’m not completely opposed to donating to my college. I liked the place a lot after all. But presenting this as a Christmas present misses the mark completely, then it seems extremely expensive, and it looks like a super inefficient way to donate, as it seems that most of your donation will be wasted on engraving a name into the stone in a somehow extremely expensive way. This feels like incredibly bad marketing in any case.

Oh, did I mention that if I wanted my transcript in a presentable way, as an official-looking document that I could, say, send to a recruiter, I would have to pay for it? It’s only like 15£, but this is still just… offputting. Like, if they’d increased the tuition fees by 15£ instead and gave everyone access by default, this wouldn’t even bother me. But this is strange and offputting, because it is an additional step of work for a large number of students, to get a document that you’d think you have an inherent right to receive, and then it’s also entirely digital and automated, so why even charge for this in the first place? I don’t get it.

What I Engaged With

I already talked at length about how much I liked my college and was active in the MCR committee work. I constantly chatted with friends in the dining hall and went to a lot of MCR events, though with a clear preference for those associated with welfare. I am reasonably sure I only missed like 2 or 3 welfare events in the entire year. One of my friends hosted a horror movie night in the MCR every few weeks, which was always fun and exposed me to a bunch of interesting movies outside of what I usually watch.

I hosted two karaoke parties during my time there: once in the MCR, and another one in the college bar. This was fun, but also had me a bit anxious. I’d done karaoke several times with friends at home in Germany, but usually I knew exactly who was coming and that everyone involved wanted to sing and had fun doing so. This wasn’t the case here, where I didn’t know if people would show up at all and if they were “brave enough” to sing in front of others - nothing is more awkward than a karaoke party where no one wants to sing. The first event had a bit of a rough start, but thanks to support from one of our social secs who organized the event with me and is a great mood maker, as well as several of my friends showing up to sing, the event was a success. For that event, I used a karaoke software called UltraStar Deluxe, that shows you whether you’re actually hitting the notes, Singstar style.

The second event went even better. Per suggestion from one of our social secs, we did it in the college bar this time, and rather than using karaoke software, we had people singing over Youtube videos, using the mics to amplify the singers' voices instead. This one had a good turnout, with up to 30 people I believe, and great vibes. We had a lot of fun up to the end, with several great performances, be it the boys who put on meme song after meme song or the girl who absolutely nailed a classic Spanish song, and many fun collective performances inbetween. One of my takeaways here was that most people don’t care all that much about the setup or whether you use a proper karaoke software. As long as you can get them to sing, it’s a success.

Choir

Beyond that, I joined a choir. I’d been in a choir for about 2 years from age 14 to 16, and I still have very fond memories of that. I was quite shy and awkward back then, but I’d always enjoyed singing and had a lot of fun there. During one of my internships away from home, only a year before I came to Oxford, a colleague suggested a local choir to me, which I’d enjoyed a lot, so I was looking for another choir as soon as I came to Oxford. Thankfully, they had a ton of choirs at the Fresher’s fair, though it turned out that most of them were auditioning. Those were a no-go for me, since I’m not that good. Even the fun choirs I’d joined before were a bit difficult to me, since I had no musical training and couldn’t sing a song from its music sheet. That significantly reduced the number of choices, essentially leaving me with the Oxford Singers and the Oxford City Singers, as well as a lot of confusion.

You see, the names were so similar that I kept mixing up these 2 choirs. At the start, I even thought they were the same choir. But the Oxford City Singers' website listed a different location for their rehearsals than the Oxford Singers' brochure, which I’d picked up at the fair, but they were both at the same time! Which made sense, because they were the same choir right? Hah. In any case, I picked the Oxford Singers' rehearsal location effectively at random, showed up to the first rehearsal and… no one was there. Perhaps 5 minutes past the supposed start of the rehearsal, I got up to leave, and I would have gone to the other location (Oxford City Singers' rehearsal) the next week, if the Oxford Singers' choir president didn’t show up that exact moment. They came by to check the room that evening and told me that rehearsals were only starting next week. This was an insane coincidence, but a very lucky one, as I spent a fun year at the choir.

While they were there at the Freshers' fair to recruit new people, they were not only a student choir, but rather open to anyone. As a result, out of the 10 or so regular members, only 3 (including me) were students, though some of them were former students who stayed to work in Oxford. It was a very diverse group of fun people, and I liked all of them. We had a good song selection, including classic old British songs like O Danny Boy, songs from musicals such as Oliver and Chicago, as well as many more old and new pop songs, including Abba’s Dancing Queen, Keane’s Somewhere Only We Know and How Far I’ll Go from Moana. I’m particularly proud of our performance of Lady Gaga’s Bad Romance, which I believe we performed well, but it’s also an absolute banger of an arrangement. I’d love to share it here, but unfortunately there’s no public recording.

The choir became much more active, socially, as time went on. In first term, there wasn’t all that much time to socialize, since we were singing throughout most of the 2h rehearsals every week. We started going to pubs sometimes after rehearsals, met up on weekends to watch the musicals whose songs we were singing, did a potluck once during summer break, and I was able to borrow our MCR to do karaoke with the choir there twice. It was a fun group, and they’re one of those I’ll miss from my time there.

Japanese Society

This is another society I signed up for at the Fresher’s fair. My main way of engaging with it were the conversation classes they offered: three classes for beginner, intermediae and advanced Japanese learners meeting on Saturday mornings. I was a bit unsure about this at first, but ended up joining the advanced class which was a good fit for my skill level. Most people there were a bit better than me, in both speaking and listening ability, but that is exactly the perfect environment to improve. As the name implies, the focus here was on conversation, which was great since that was by far my weakest area. My fellow Japanese learners in that class were mostly people studying Japanese Studies, but there were also a few others. The numbers quickly thinned out until, towards the end of the term, it was only me, three other guys as well as a high school kid from the area.

The conversation class itself wasn’t that exciting, but it was a good opportunity to get to speak a bit. Our teacher (a Japanese student) was a big fan of temples and shrines apparently, and we spent 4 or 5 weeks only talking about those. I certainly learned interesting things there, but man, I’ve had enough of temples and shrines for a while.

Perhaps the more interesting part of this were my fellow Japanese learners. They invited me to go out for lunch after class a couple times, but I declined almost every time during first term since I was too worried about my finances. All of this got better in second term, so we went out to eat together at the market in city centre a few times, and started frequenting Edamame, the Japanese restaurant I praised earlier, as well as several Tapioka stores together. Eventually, we started our own “Benkyoukai” (Japanese study session) on Mondays, which evolved into a lot more.

We usually cooked Japanese food, e.g. Tonkatsu, Karaage, Shougayaki and similar, and then watched a Japanese TV show (Terrace House) together. While the show was fun to watch on its own, we also stopped frequently to discuss words or terms and their usage, and we frequently chatted with each other in a mix of English and Japanese. All of this was a great environment to learn Japanese, especially the speaking and listening part that I wasn’t so great at. It is difficult to judge how much better my Japanese got through this, or even during the entire year there in general, but at the very least I can say I spent a lot of time practicing in new ways and had a lot of fun.

We did our cooking at one of the guy’s student accommodation in St. John’s, because it is quite central for all of us and the kitchen was pretty nice. They also had a music room there, complete with mics etc. and my friend living at St. John’s just so happened to be in a band and own some equipment such as a mixer. Before too long, karaoke parties in the St. John’s music room had become a regular part of our activities. This was a lot of fun, and I discovered a lot of great music this way. We were all into Japanese music, and thus ended up singing almost exclusively Japanese songs. This was a cool opportunity for me to sing all those songs I’d always wanted to sing, but rarely could because my friends weren’t that into Japanese music (if at all). This was the point where my knowledge of Japanese music finally escaped the “You just watch anime right?"-level.

The guy who’s in a band frequently encouraged me to learn difficult songs with him, and while I wasn’t able to do nearly all of them due to time constraints, I did fully learn and memorize LOSER by Kenshi Yonezu and nqrse’s rap in Taiyoukei Desuko. Both took a lot of work to learn, but the result is worth it. It is incredibly satisfying to be able to sing those cleanly, and singing them together with him was especially fun. At some point, he gave a concert with his band that I went to, and it was so fun to see him perform several Japanese songs I knew. He also made me realize that I’m not nearly as good at singing as I thought I was. Hearing someone absolutely nail the “gyakkouuuuuu” from the end of the chorus of Ado’s Backlight was beyond cathartic.

At some point during my studies, Japanese singer Ado did her first world tour, and 3 of us 4 decided to go to her concert in London together. This was my first time going to a concert, and I was a bit worried about this since I’m very noise sensitive. I even ordered concert-earplugs before this to be safe. Surprisingly, the noise wasn’t much of a problem, but the experience was still mixed for me. We lined up about 6h before the concert, to ensure we got a front-row place once the gates opened. The others lined up even earlier, but allowed me to join late since I had a job interview that morning. In any case, my experience that day consisted mostly of sitting on the pavement in front of the entrance for many hours. Once the concert began, it quickly got very hot in there and I found myself very thirsty within the first 20mins, but we weren’t allowed to bring drinks in there. You cannot imagine the relief when I finally got some water after the concert ended 3h later. The concert itself wasn’t bad by any means, but I found it difficult to make out her voice over the instrumentals, and overall felt like I got a better experience listening to her music via headphones when I was alone in my room. That said, I was super pumped when the intro to Ashura-chan started playing, and hearing the entire audience go “Ussee, Ussee, Usseewa” together during the chorus of Usseewa had this amazing feeling of energy that is impossible to explain.

The Japanese society had a bunch of events going on throughout each term, though I didn’t go to that many of them. The welcome drinks at the start of every term were a fun opportunity to meet people, though they were the kind of event that I’m usually not that comfortable at. Still, I had a great time, especially at the last one, where the society had invited a group of Japanese exchange students who came to Oxford to study English, and they were super fun to chat with.

The society hosted a sushi making event which was a lot of fun, and also just delicious. I’m particularly happy about how well my rolls turned out. One of the other participants there even asked if she could take photos of mine to put on her Instagram. Other events included origami making, as well as a career event which I sort of accidentally stumbled into, despite the fact that it was meant for Japanese students. It was still good to see that I was able to follow the quick talk of that woman presenting for a Japanese recruitment agency. Finally, they organized a movie night where we watched Asakusa Kid - super interesting movie, and it was fun to find out afterwards that this was about the life of a real person, namely Takeshi from Takeshi’s Castle of all things.

The best part of joining the Japanese society however, was the opportunity to meet three wonderful people there, who came to shape my experience in Oxford and whom I became great friends with. For now, we are keeping in touch by doing a video call every Sunday, but I am convinced I will meet them in Japan at some point.

Teaching German

I’d been visiting the Japanese society’s conversation classes from the start to the very end of my year there. I was also signed up to the German society’s newsletter, though I didn’t engage with them beyond going to their start-of-term “Kaffee & Kuchen” (coffee & cake) event. I did see them mentioning that they offer language classes, and are still looking for more volunteer teachers though. I thought it would be nice to give something back after attending Japanese conversation classes at the time, so I messaged them at the start of my 2nd term, to ask whether I could help out. I didn’t hear back at all until right before the start of 3rd term: turns out they didn’t have anyone to manage the classes, so no one was even monitoring the mail address I’d messaged, until the new guy showed up. He’d been wanting to teach German classes since the start of the year apparently, and since there was no one to organize that, he eventually took over the job himself for the 3rd term. I find this quite impressive, especially since he was there for only a year as an exchange student.

In any case, there was another person who’d volunteered to be a teacher, so we each got to teach our own class alone, following the classic beginner-intermediate-advanced scheme. I got the intermediate class and was assigned perhaps 6 or 7 students. Now I’ll have to say I don’t have any experience at all teaching languages, though I do naturally have experience learning languages, especially from my Japanese studies. Our chief teacher had shared with me a bunch of materials provided by a German government project, designed to teach German to people of all levels. I looked at these but found them a bit too school-ish for my tastes. I’d seen the conversation classes at the Japanese society, which were mostly a simple get-together with a native speaker, to chat with each other in Japanese under the supervision of a native speaker who could help out with vocabulary, grammar and cultural questions. This was relatively low effort on the part of the teacher, but still made for a very flexible learning environment. It was all we needed honestly.

So, I didn’t think too much of the materials I’d been sent, and would have preferred doing it JapSoc-style. That was less work to prepare materials for me of course, which would be preferable since I was busy with my thesis work and exams at that time. But I also believe the average uni student wouldn’t appreciate more school-ish exercises. In any case, I prepared some materials just in case and decided to see how it goes for the first session.

The answer is: not that well. I had a positive first impression of my students, they all seemed to be nice people, including a very charming elderly lady who must have been over 70. Moreover, their German was quite impressive already, and I could have normal conversations with them, without paying too much attention to my choice of words. This was a good sign I thought, as this should make it easy to turn this into a conversation class like in the Japanese society. But the class was for them of course, so I asked them first about their preferences for the class and how or what they would like to study. While they gave me little to work with, they were at least clear about wanting to train their speaking ability especially, since they could do everything else on their own in their free time. Nice, I thought. Time to copy the JapSoc’s formula.

Unfortunately, I had overestimated their ability to have normal conversations with each other. Any attempt by me to start a conversation quickly devolved into a rather drawn out question-answer game. I asked “So what do you guys study?”, and they took turns answering the question, same with my other attempts. They barely engaged with each other at all, effectively leaving me to manage the entire conversation. But if you know me, you’ll know that this is far from my specialty. As a result, a decent part of the classes turned into me asking a question, followed by the students answering them sequentially. I tried to make it more personal by digging deeper, asking for more details whenever they gave me something to work with, but they tended to be quite reluctant to answer in more than a single sentence. This had me stumped, and very stressed out. I had a hard time listening to their answers because I was busy thinking of the next thing to ask them, in order to not let awkward silence encroach on the class. They were all quite reserved and appeared almost disinterested, despite them seeming like nice people, and essentially refused to talk to each other at all.

All the while, I knew I was delivering a bad teaching experience, but I didn’t know how to do better. I’m bad at keeping a group conversation going, they told me what they wanted to do but didn’t seem to want to engage with this, and worst of all, no one gave me any feedback. When I ran out of ideas to keep the conversation going, I pulled out the teaching materials I’d prepared. I filtered it down to the exercises that involved speaking, which wasn’t much, but barely sufficed to carry me to the end of the class. Then I asked them for feedback. I made it very clear that I was here for them, to teach them German, and that I was happy to consider any wish they had, change the teaching format however they wanted etc. I got no reply at all. They told me it’s fine this way.

This kept going with participant numbers slowly dwindling, until I taught a single person in weeks 6 - 8. After the first class, I made sure to prepare more teaching materials so I could at least make it through a single lesson without running out of stuff to do, and that made it a bit easier. But even so, every class was a stressful experience, I dreaded Saturdays a bit, tried to just get through it when the time arrived and was relieved when it was over.

Overall, this was a very unsatisfying experience. I felt like I didn’t do a good job as a teacher, despite best intentions. At the same time, I’d learned nothing through this. I have no idea what I did wrong, and how to do better next time. In fact, I’m not even that sure my students were unhappy with this. I always made it clear I was open for and happy about any and all feedback, and no one ever complained about anything. The fact that less students showed up as time went on is only natural, especially as all of them approached exams. Still, this was confusing. One of my friends from the Japanese society suggested that our conversations classes only went so well, because we were all friends and naturally enjoyed chatting with each other. One of them was quite well connected within the society and thus knew most of the teachers decently well.

I also asked my fellow teachers how their classes feel like. The advanced class' teacher mentioned that her students were super motivated, and begged her for homework. She put a lot of work into the class because of this, which I found quite impressive, but at the point where she mentioned “super motivated” it was clear that her experience probably couldn’t apply to my class. The beginner class' teacher was doing mostly very school-ish rote-teaching, since there wasn’t much else you could do with beginners. I remain confused and disappointed by this experience to this day. I don’t think I’ll be teaching a German class again any time soon, or at least not alone without another person for emotional support. I still hope I managed to do something good for some of my students, but I honestly don’t know.

Open Source Contributions

Something essentially unrelated that accompanied me through this year, we’re a lot of open source contributions. The first time that I felt like I actually knew programming to the point I could do something useful with it, was in 8th grade when I programmed a simple commandline birthday calendar script in Python. One of the cool things about programming is that you can reasonably pursue this in your free time, in contrast to almost all other sciences. After this first meaningful project, I’d always felt like I wanted to do something genuinely useful, not just to me. The most obvious choice here was contributing to open source projects. How about making my favorite browser Firefox better, or add a feature I’m missing to the Signal messenger? But for a long, long time I failed to do so, because I didn’t know how to start. All of this changed significantly during my time at Oxford.

I’ve already been contributing a bit to two of the tools commonly used in the UltraStar karaoke software ecosystem. But this intensified during my time in Oxford, I became a maintainer in both of these projects and started contributing regularly to the UltraStar Deluxe karaoke software itself, which I’d frequently used myself for karaoke parties with my friends. It’s a very old project, and it felt like I could make a noticeable difference, even with simple bugfixes. Seeing that the software is now in a noticeably better state than before I came to Oxford, is great to see.

While somewhat less relevant to me directly, I started contributing to LLVM, the mother-project of the Clang compiler. This is well known in the software engineering sphere, to the point that I could genuinely brag with this. It’s not like I’m doing something incredible, I’m not even working on the compiler but only on the Clang-Python bindings. That is nothing but a small side-project within LLVM, which is already 15 years old and has been neglected for most of its lifetime. I initially planned to contribute only type hints here, as I was dearly missing those when I was using the bindings for a task during my second internship. But the guy reviewing my patches encouraged me to look at other things as well, and so I’ve added my typehints, simplified the code, fixed several long-standing bugs, improved test coverage and much more, in about 20 pull requests to date. I got write access to the project, have become code owner for the bindings, and am now reviewing patches several times a month. I’m super happy about this.

But perhaps the most important thing here, is that I’ve become comfortable engaging with new projects, finding my own niche and contributing in whatever way I can. I’ve opened pull requests on MXE, MyPy, PyTorch, several parts of LLVM, KDE etc. in the past year. I feel like I’m finally on the path to becoming the engineer I want to be, and that’s great.

Other Activities

A couple more things, that don’t quite deserve their own heading, but still accompanied me for an extended amount of time during my stay.

I frequently cooked with friends, despite the setting (mostly small uni accommodations) being not exactly conducive to this. It was a lot of fun, and I managed to gather a lot of experience especially with Japanese dishes. I miss the food we made as it was mostly super good, and the lack of rice cookers in my vicinity in Germany saddens me. I’d like to present a picture here of one time we made Japanese-style Omurice. This was the most attractive omelette I’ve seen in my life:

The German society’s Kaffee & Kuchen events made me think of Kalter Hund, a German cake consisting of layered biscuits alternating with a molten chocolate mass, and I got a real craving for it. So I made Kalter Hund for the first time in my life, and then happened to make it a couple more times, for several events in the MCR, our choir potluck and karaoke etc. It was described by most as either “too heavy” or “making me rediscover my animalistic instincts” with almost nothing inbetween. Other quotes included “Almost killed me, but it was worth it” and “Lock it away in your room […] I might not be able to help myself”. I see this as an absolute win. This was also one of the few recipes in my life I made so many times that I had the opportunity to change the ratios and optimize the recipe, so I felt like I got better and better at this the more I made it. Very satisfying.

Towards the end of my third term, I found out that some of my college friends were going bouldering frequently, so I joined them after that. I love bouldering but I didn’t get to work out a lot during my stay in Oxford, so this was a welcome distraction. Our local climbing gym had a slackline which we frequently used, and I realized once more that my sense of balance is quite bad. Even balancing on one foot at the starting position is extremely challenging to me.

Our MCR has its own academic journal, completely run by MCR members and publishing mostly papers and other submissions by MCR members. I was roped into this by two friends and thus ended up becoming an editor for this year’s edition. There isn’t that much else to say here, we had a couple workshops on editing etc. and I reviewed two papers. Publication is coming up at the end of this year, though almost everything except editing is handled by our editor-in-chief whom I do not envy for their extremely busy position.

There are multiple arms of the Thames running through Oxford, and somehow “punting” has established itself as a typical Oxford thing. This is gondolas maneuvered with a long pole, think Venician canals. I did this only towards the end of my studies once summer reared its ugly head, twice in total. It was fun, but navigating like this is genuinely hard. It is almost impossible to move in a straight line and moving the 5m pole is slow and takes quite a bit of strength. Cool activtiy though, and members of college can do this for free every single day in summer. That’s surprisingly generous by British standards.

I was planning a movie night with friends in our MCR at some point, when a friend of a friend came and proposed we go watch the norther lights instead. What?

Well, we did in fact get to see the northern lights on a warm summer night in Oxford. We ended up wandering into the center of Port Meadow, a big park west of Oxford. It took a while for them to become visible at all, and there wasn’t that much to see with the naked eye, but it was still kinda cool. Somehow, people got way better pictures with their phone cameras than what you could see in real life, and they were going crazy over them in various group chats. It felt a bit fake to me though, and the lack of “realistic” pictures means that everyone who slept through this probably thinks the sky was deep purple that night.

While we were waiting for the lights to appear, we had this slightly creepy event where a weird scratching sound was very slowly approaching us over the course of several minutes. At some point we realized the dark hunks we could barely make out behind us were wild (?) horses, and we were hearing them feeding on the grass from the meadow as they were slowly approaching us.

I’ve always wanted to learn to do handstands, and a friend from college motivated me to practice this. Fortunately, I had enough space in my room, and started doing handstands multiple times a day, with my back against the wall. But somehow, even after months of doing this, I never got to the point that I can keep my balance after pushing myself away from the wall. I don’t know what I’m doing wrong, but even after doing hundreds of handstands over the past 12 months, I still cannot do it. Also, doing handstands makes me sneeze for some reason. Huh.

Towards the end of my stay, I had a weekly timetable much fuller than ever in my life before, and similarly, I had never socialized as consistently much as I did during my time in Oxford. Monday evenings were for Japanese cooking & studying with my friends from the Japanese society. Tuesdays were there for playing Stardew Valley online with a friend from Hamburg, and later for bouldering. On Thursdays I went to the MCR’s “Shut up and write” events in the study room in the mornings and to choir rehearsal in the evenings. Fridays frequently had movie nights hosted by my friend, or other events. On Saturday mornings, I had Japanese conversation class followed by lunch with my friends and German class after that. On Sundays, I always went for brunch with my friends in the college dining hall, and we would chat until hall closes. In the afternoon or evening, I would meet online with friends from home to talk about the anime we watched together weekly.

What I Learned

There are a lot of things I take away from my year in Oxford, and here I want to talk in more detail about some of them. This section will be mostly very personal and introspective.

On English

For one thing, I learned that I’m not as good at English as I thought. Perhaps this should have been obvious when I got barely enough points in TOEFL to be admitted to Oxford, but I certainly walk out of this with less confidence in my English ability. And that’s not because my English got worse during my stay in Oxford (arrrr the British): rationally speaking it probably got better, but I was surrounded by people whose first language, for the most part, wasn’t English, and who still spoke notably better than me. I have always been a bad speaker, even in German, meaning I had an even harder time in English. None of this was a serious problem of course, I’m still fluent and can communicate everything I need to. Yet, I still felt like I wasn’t expressing myself well, clumsily even, that I often struggled to come up with a word I needed even though I technically knew it etc. It wasn’t unusual for me to ask “what’s the word for <description> again”?

I also realized just how bad my pronouciation is. I was frequently asked “Are you German?” within two sentences of talking to someone new. One could clearly tell that I had a German accent still, and that bothers me. I don’t mind accents in general, but seeing how I was still so obviously imperfect at even these baisc parts of the language after using it for over 15 years, is a bit frustrating. Contrast this to a non-native speaker I met, who learns accents for fun, which I find quite impressive when I can’t even turn off my own. My German friends joked before I went to Oxford that I might return, speaking of “bo’els of wa’er”, and I am, for better or worse, far from that.

On the less negative side, during my stay in Oxford, I sometimes had this random realization as I was biking to college in the morning, that I was entering into a day that would play out entirely in English. This is obvious, but sometimes I’m still amazed at the thought that I was living in a different country, speaking a non-native language, and was getting by like that just fine. Because the English level among students in Oxford is so high, one frequently runs into the question of whether to speak English or German with other German students. I ended up speaking German with most of them, but I also spent hours chatting with our MCR president when we were alone in the study room, and we never once exchanged a full sentence in German. In fact, when she dropped “Schwarzbrot” during a conversation about bread once, I got so confused by this foreign word that I didn’t understand and asked her to repeat herself twice.

About halfway through my studies, I finally resolved to add an English/German vocabulary flashcard deck to my Anki collection. I should have done this earlier to be honest since I already missed a lot of learning opportunities that way, but it was still a good thing. I used this both ways, from English to German for words I encountered but didn’t understand (e.g. squeamish, pernicious, recalcitrant) and from German to English, often for words I technically knew, but couldn’t recall when I needed them (e.g. curls, ladle, nuclear fission). I should have put more effort into this, since after a few months the deck still sits at 50 cards, but it’s something.

I learned a bunch of things about pronounciation and meanings of words that I got plain wrong before then. Place names are especially bad in the UK: Magdalen college is pronounced “Modd-lenn”, Bicester is pronounced “Biss-ter”, Reading is pronounced “Redding” etc. I learned, only a month before returning to Germany, that “pudding” means something entirely different in the UK. We use “pudding” as a loanword in German, which seems to be close to the American “pudding”, i.e. a creamy, custard-like dessert. In the UK, pudding can refer to… sort of everything it seems? Apparently it just means dessert, except when it does not, similarly to how “pie” is used for many savoury dishes there.

On German

An important takeaway for me was, that most people don’t think negatively of Germany or the German language. I didn’t think so either for most of my life, since I’d never seen much negative commentary about Germans on the internet, and I knew that the German passport was quite useful. But I fell in with a group of rather bad “friends” from other European countries some time before Oxford, and they took every chance to somehow make fun of the country and the language, and to tell me how no one likes Germans. While I stopped talking to them a few months before coming to Oxford, they had pretty much convinced me by that time. As such, I was positively surprised after getting there. I believe about a quarter of the people I met in my first week or so in Oxford could even speak German to a decent level, if not fluently.

One of my flatmates had lived in Germany for a while and was fluent. One of my friends in college was studying literature with a focus on German works, and another one was studying the economics and history of Germany. They’d asked me at some point to help them decipher a piece of old German handwriting from the 19th century I believe, but I was of no help, because they were better at it than me! I met the Japanese society’s president at their first event, and chatted with her in German for a bit, since she had been studying the language for a while. Our MCR president was German, and one of the social secs on the committee could speak pretty fluent German. In the CS department, I often hung out with a group of Germans, and there were a Dutch and Ukrainian guy who often listened in on our German conversations, and apparently didn’t have much trouble following along. And obviously, I was teaching German to a bunch of people in my classes. There’s a very high chance on any given day in Oxford that you’ll randomly hear spoken German from people passing you by. Germans are very present in Oxford, and there are a lot of people learning the language.

The latter part especially surprises me. The fact that people learn German seems strange to me, like, why would anyone do that? Maybe it’s that, having grown up in Germany, I naturally knew almost only natives, so there’s barely anyone in my vicinity who did learn the language. German doesn’t strike me as a very interesting language, but that’s difficult to judge as someone who had that skill his entire life. One of my friends even told me that they really like the sound of spoken German. I am far from nationalistic, but that still made me happy.

Socializing

I met a ton of people in Oxford. I tried to estimate how many people I met on my flight back to Germany during the first term break in December, and the best approach I came up with, was counting new contacts on my phone. I counted more than 70, and that was after only my first two months there. Of course not all of these were friends and I probably didn’t talk again to most of these, but this is still far, far more than at any other time in my life before this.

I was in a somewhat comparable situation when I went to Aachen for my undergrad: a university city where, statistically speaking, every fourth person was a student and the majority had moved there to study. As a result, most people were very open as they tried to make new connections and find friends there. Oxford is the same, but turned up to 11. People didn’t just move cities, but countries in this case. I didn’t meet a single Oxford native, but instead incredibly interesting people from all over the world, who study all sorts of things and have all sorts of diverse backgrounds and hobbies. This made it incredibly easy to walk up to new people, introduce yourself and have an interesting conversation at events.

Moreover, I kept getting to know new people, or became friends with people I’d sort of known for months already, as time went on, which was especially surprising towards the end. My course was 11 months long, but most other courses are shorter than that. I had friends who finished studying and left the city two months before my course ended, and I went to a friend’s graduation before I even handed in my thesis work. This meant that I had a much more lonely last month in Oxford. But with most of my good friends already gone, I ended up making almost completely new connections instead.

There’s this guy from my program who I’d chatted with on Discord about our course selection, exams etc. a lot, but barely talked to in real life. Then one day in July I ran into him by chance in the city center, and we ended up meeting to cook together several times after that and also to do karaoke once. I had a friend in college who I ended up simply spending a lot more time with than before, cooking together at least once a week, watching a bunch of movies and documentaries together etc. There was someone I met at the Japanese society’s conversation classes, but only 2 or 3 times since they didn’t participate most of the time. I then randomly ran into them at my accommodation in my last month in Oxford. We’d lived in the same accommodation for the entire year without noticing! After that, we cooked together and did karaoke once. There’s more stories like this. I still made new friends in my last month at Oxford, when I was expecting it to be sort of sad and lonely.

That said, it was still sad and lonely in other ways. Most students have either finished their courses already or left over the summer break because they get kicked out of their student accommodation. They were replaced by tourists, and the city felt like a completely different place now. I would say I have a sort of nostalgic attachment to Oxford, but that isn’t so much bound to the place itself. What made it “Oxford” for me were the people I met there, and with most of them gone, it suddenly felt foreign. This wasn’t the same place anymore.

As I said earlier, I’ve never socialized as much in my life as I did during my time at Oxford. That doesn’t mean that I’m an unsocial person. I’m probably at my happiest when I hang out with people I really like, but I’m also completely fine being alone. In any case, this amount of socializing, and especially interacting with a large and diverse set of people and social groups, allowed me to make many interesting observations, with regards to how I interact with people.

There is a popular metaphor about introversion and extroversion that goes something like this: “Extroverted people recharge their batteries by socializing. Introverts discharge their battiers that way, and recharge through alone-time.” I’ve always found this interesting, but could never relate to it. I see myself as an introvert, but I don’t always find being with people to be tiring, and with the right people it can be the most energizing thing I know. Out of curiosity, I asked many of my friends in Oxford whether they can relate to this metaphor. To my great surprise, almost every single one of them affirmed it. Going from hanging out with friends 10 to 15 times a week in Oxford, to maybe 2 or 3 times a week back at home, felt a bit strange to me, but it wasn’t a problem at all.

I would say that, generally speaking, I have a lot of friends. I am still well in touch with more than 5 of those from my school days whom I’ve known for well over 10 years. Despite my undergrad time being wrecked by Covid, I am still friends with 10 or so people from Aachen, checking up on them every once in a while and visiting when I get the chance. And then I met a ton of cool people and made many new friends in Oxford. I need to spend a lot of time with someone to get the feeling that I know them well and feel very comfortable around them. If possible, I’d like to get to that level with every friend I make, but this makes every friendship a large time investment.

My brother has given me a lot of career and life advice over the past few years, and one of those was to find just 3 to 7 people to treasure as your best friends, because more than that would be unrealistic and unnecessary. My mother often told me that friendships are temporary, that often you fall out of touch with someone because life moves on, and not to be too sad about that. I’ve always disagreed with them on this. Being friends with someone means a lot, and I would like to keep someone whom I consider a friend, for as long as possible. The idea of losing friends, in the absence of a serious issue, or somehow deciding not to invest more time into someone, seemed sad and alien to me. But over the past year, I think I’ve come to understand what my mother and my brother meant by this.

Becoming close friends with someone requires a lot of time, and I think I already have a lot of friends. That number completely exploded in Oxford though, and quickly made me realize that there’s a limit to how many friendships I can maintain. Well, that much should be obvious, so perhaps the more useful insight was that I can realistically reach this limit. It quickly became clear in Oxford that I met too many cool people to reasonably have a good friendship with and that’s okay! My time there made me accept that I’ll have to make conscious decisions about who I’m spending my time with, because otherwise my limited resources will make the decision for me. Similarly I’ve come to terms with the idea that some people will be only temporary friends, since we’re not close enough to stay in touch. And that’s okay too! It doesn’t make the time we spent together any less meaningful.

As an example, there’s this one person in my college who was in the same accommodation. We were never too close, but we met by chance at various events or sometimes in the dining hall and always had good chats. They were one of the few who stayed in Oxford almost as long as me, and as “fellow loners” we sort of kept each other company at the end, going out for a coffee or a walk together once a week. I saw them again at our graduation not too long ago and that was nice, but there is a decent chance that we’ll never talk to each other again. I don’t consider them a close friend, but I still like them and we had fun together. I have no regrets.

One part of this is the realization that a lot of friendships have some kind of “situationality” to them. I mentioned before how the dining hall was a social hotspot for me, but that’s not just for fun: it enabled a lot of social interactions that would simply not have taken place otherwise. I met some of my closest friends in the dining hall on a weekly, if not daily, basis, but some of them I might not have seen at all if the dining hall didn’t exist. On the flipside, there are people who I got along with really well in certain contexts, but as soon as that context disappeared, our relationship ended as well.

Similarly, there are people whom I like a lot, but the idea of spending, say, a long car ride alone with them is horrifying because I’m not confident I could keep up a conversation for more than half an hour without it getting very awkward. I used to think of this as the main thing I look for in a friend: the ability to chat with them for hours with no effort. The latter part is important, because sometimes I can keep up a long converation with someone, but it takes a lot of conscious effort to actively search for topics etc. Then there are others who I can chat with for 6 hours straight, and I barely notice because it goes so well and time flies by. But this is far from the only thing that matters, and I found that I have some great friends who this doesn’t apply to, and instead I appreciate them for very different reasons.

I believe I generally have a pretty accurate model of the social environment I’m in, i.e. I know who’s friends with whom, what role each person takes on in a given group, how they are perceived by others etc. The one strange thing about this, is the fact that I am not a part of my own models. I have no idea how other people see me beyond what they explicitly tell me. Furthermore, since I’m not a part of these models, the relationships of other people don’t depend on me in my own models. In other words, in my imagination, these groups I’m a part of would have existed just the same if I wasn’t there. As an extension, it feels like the way I act doesn’t affect the people around me. Now all of this is, of course, factually wrong, but I usually don’t notice unless the consequences are so obvious that they are clearly observable and attributable to me, or my friends explicitly tell me how they perceive me or the dynamics of a group I am part of.

As a basic example, I went bouldering on a weekly basis with friends in Hamburg. Months after I left for Oxford, I’d talked to them about this and was quite surprised to find out that they’d completely stopped going right after I’d left. Due to the absence of any other factors that could explain this, the only possible conclusion is that somehow I was the one motivating them to go bouldering all this time.

Similarly, it seems that I somehow have a very good reputation within our MCR and I have no idea why. That’s judging from some things people have told me directly, as well as what little I garnered from how people act towards me. When I joined the MCR committee towards the end of the year, another committee member told me that I was “popular”. Several people have told me, independently from each other, that they see me as a “cheerful” or “very sweet person”. Now some of this I can attribute to the overall positive vibes in our MCR, but not all of it.

Someone from college asked me if I could help one of their friends, since they were struggling hard to find happy people in Oxford (lol) they needed for comparative purposes in a study they were conducting about depression. I participated in the study and was told afterwards that, based on how I answered the questions, I was indeed “one of the super happy people”.

Being described as “super happy” or “cheerful” was particularly interesting, because I don’t think anyone at home would describe me that way. When I told my mother about this, she found that quite funny, and said “they must not know you too well”. I think this is incorrect, but I can’t fault her for her conclusion. I am aware that I acted differently in Oxford, or at least within my college MCR, than I do, say, at home in Hamburg. But I never did so on purpose, this is something that happened on its own. If anything, I feel like I am simply reflecting the positive vibes of our MCR by acting in a way that is consistent with how I see our MCR members act. I didn’t think I was particularly sensitive to the culture and atmosphere around me, not to the point that I could easily, let alone automatically, adapt to it, but I guess I was wrong.

I believe I have relatively little empathy. It’s always been this way and I’ve been working on this for almost as long as I can think. I’ve made meaningful progress in that direction, but seeing how others perceived me in Oxford was nice, as it reaffirmed that feeling from the perspective of completely new people. That said, I still have a long way to go. There was a situation in a group chat I was in, where someone in college expressed their worries about certain events to a member of the group chat, and that person wasn’t sure how to help them so they asked us in the group chat for advice. I saw this message and immediately knew I should not respond to this, because this was precisely the kind of thing I am really bad at. But then no one else in the group chat responded for an hour or two, and since I felt bad for the person who asked, I eventually wrote how I would deal with this. And honestly? It was bad. I shouldn’t have sent this message. I knew from the start I shouldn’t do this and I still did, against my better judgement, which made this particularly frustrating. A couple of the other group members replied shortly after that, and they started their messages with “I would rather […]” or ignored my message completely (which I am quite thankful for) and, as expected, proceeded to give way better advice than I ever could have. There goes my reminder that I still have a lot to work on, but at least I have better intuition than I thought I did. Perhaps I should listen to it.

I often find it difficult to give compliments to people, and I want to get better at this. I made a wonderful friend in Oxford who is the exact opposite of me in that sense. Sometimes when we walked through college together and they, very naturally, dropped compliments left and right to friends and acquaintances we passed. I find this quite impressive, especially because I’m so bad at it in comparison, but they were also an inspiration for me to try harder. This is one of the reasons why I was so motivated for our Palentine’s event: it was a great opportunity to give compliments to my college friends and tell them how much I liked them. This was a sort of workaround since I’d rather tell people in person, but it’s still progress in my opinion. It’s particularly hard for me to compliment women, or to compliment people on their looks. Complimenting women on their looks is almost impossible then. These are just part of the many weird and frustrating anxieties I have.

I’ve met a surprising number of people with mental health or other psychological issues at Oxford, and some of them were very open to talk about this in ways that I’d rarely seen before. This was refreshing and I learned a lot from that. For one thing, it made me realize just how prevalent these issues are. My friends in Oxford were very open in talking about how they manage their social life, which was great. The latter helped me accept that going to events and meeting people can be difficult and takes effort for many of us. This should be obvious, but I feel like years of exposure to (social) media has kept reinforcing this image of introversion being something you have to fix. I am supposed to go to events. I should enjoy big parties. And if I don’t, then there’s something wrong with me I need to work on. But this isn’t true at all. I had several friends who sometimes dropped out of events on short notice because they’d realized they’d planned more social time for that day than they could handle. I had a friend telling me several times that they couldn’t e.g. meet me on Saturday evening since they’d already done something with friends that morning and it would be, I quote, “more than my tiny introvert brain can handle”. I think this openness is wonderful, because it allows us to have conversations about our personal limitations, and to not feel bad about them.

Some people see intrusive, disturbing images in their head, say, of them hurting their family or friends. I’ve had this for a very long time. I’ve barely given it any conscious thought, and simply accepted it as another one my strange issues that I would never talk about with anyone. It appears at random and I ignore it as best as I can, but it still scared me. It made me wonder whether I’m a threat to people who are important to me. I had a conversation with someone in Oxford, where they ended up telling me that they have this exact issue. It is apparently a part of OCD for many, and having this doesn’t mean that you’re dangerous, or that you want any of the things you see in those images. Can you imagine what a relief it was to find out that you’re not alone with this, and that it doesn’t say anything about you as person? It’s like this worry I’ve been carrying around with me for longer than I can remember just disappeared through a chat with a friend.

Memory

My memory is bad. Well, that’s not quite accurate, but I often forget things. Especially short-term memory, but I’m great at remembering things like numbers, equations, many kinds of facts. Now, being a bit forgetful at times isn’t that weird, but it is the extent worries me.

A somewhat recent event illustrated that pretty well. There’s a bouldering gym close to where I used to live in Hamburg, and I often went there several times a week since it opened a couple years ago. I started dragging my friends along to go bouldering there some time last year, and a couple months ago I went once more with a friend. That time, I noticed a hole in one of the walls, with a circular shape and a diameter of about 10cm. I looked into the hole and found that they put a little wall made of lego inside the hole, complete with a little figurine climbing the wall. It’s such a small thing, but I found it quite cute and memorable. I immediately called out to my friend to show it to him. He looked into the hole, then looked at me as if I was stupid and said “You already showed this to me last week.”

Now this is both funny and… scary? I’ve been to this bouldering gym over a hundred times, I’ve never noticed this before and was so excited about it I had to tell my friend right away. Yet I completely, perfectly forgot this existed to the point that I thought I’d discovered something completely new, within seven days. No matter how I look at it, this is not something that should happen to a human with a healthy brain.

I recently read an interview with a dementia patient, and the level to which I related to it was alarming. It wasn’t a late stage patient, obviously, but all the small things you got used to, those patterns of behavior you adopt to counter your weak memory, all of it describes my life. The frantic, obsessive note taking, writing everything down into calendars and todo lists and other files because you know if you don’t you’ll mess up. Walking back and forth between two rooms three times in a row, because by the time you reach the other room you already forgot what you were supposed to do there. The sudden feeling of dread when you realize out of nowhere that you forgot something but you have no idea what. This and so much more.

I don’t seriously struggle with managing my everyday life of course, and I would even say I function pretty well, but I am genuinely unsure whether I could lead a normal life without my todo list, calendar etc. at this point. I’m not just “a bit forgetful” like a sidenote you’d see in the description for a video game character. This affects various parts of my life, and I’m not sure I’ve even noticed all of them. My mother sometimes complains that she feels like I don’t listen to her or care about her because I forget the things she tells me. This is extremely frustrating because it makes me feel bad, but there’s nothing I can do about it. I’m already trying so hard not to forget things, I don’t know what else I could possibly still do. At some point I started writing down all sorts of facts about my friends and family members, such as what their job/study situations look like at the moment, what kinds of food they do or don’t like etc. I eventually stopped, because it felt weird and it’s not practically possible to maintain.

As a result, I am easy to gaslight. I’ve been in situations before where I thought my friends were telling me random shit that supposedly happened to make fun of me. Sometimes they do, but usually they don’t, and I cannot tell the difference sometimes! Mind you, we are speaking of stories about situations where I was supposed to be physically present, so if those things are true, they’re supposed to be somewhere in my memory. This is not much of a problem usually because I have great friends and they would never take advantage of this. But I’ve already been through one toxic relationship, with someone I can’t describe as anything but a genuinely bad person. It didn’t last very long so I don’t think she ever noticed, but the idea of someone abusing this weakness is terrifying.

This is ultimately one of the reasons why I’m writing this article, and perhaps the main reason it is so excessively long. My year in Oxford was incredibly dense and so full of small and big moments that I want to remember for years to come. But knowing myself, the vast majority of these I will simply forget. In just a year or two from now, there’ll be no other way for me to relive many of those moments, than to reread what I’ve written down in this article, or ask my friends to tell me about the things they remember. This is not only about the facts, since I tend to remember facts pretty well, but even more so about capturing how I felt at specific moments in my life or about certain situations.

The latter is important, because I noticed that hindsight is extremely inaccurate when it comes to judging how I felt about certain things. This is one of the things I only finally realized this year, specifically through the Ado concert I mentioned earlier. As I described above, my experience was mixed at best, to the point that I distinctly remember thinking, halfway through the concert, that I don’t think I’ll be going to a concert again after this. I wasn’t exactly disappointed, but I realized this just isn’t really for me, which wasn’t all that unexpected. Skip ahead 6 months, and if you asked me how I feel about the concert experience looking back now, I could summarize it as “it was fun”. My emotional memory of the concert is practically entirely positive, despite the fact that I factually know I wasn’t all that happy!

I never warmed up with diaries much, so writing my blog like this is something I do for myself in many ways, including being a sort of emotional inventory for things I’ve experienced, precisely so I do not forget how I felt in certain situations. Doing that in this blog format motivates me to write in a more digestable way (he said, as he typed the 35017th word of the article) so it is easier for me to go back, reread and reflect on my experiences. I sincerely apologize to anyone reading this, I still think parts of this article are interesting, even for outsiders, but I’m well past the point where I could recommend anyone read the entirety of it…

Noise

I always knew that I was sensitive to noise, but after a long, long time Oxford helped me realize that this goes beyond just “being sensitive”. This is an important realization, because I assumed I was at best mildly different to the people around me, but that’s not the case and it explains a lot of things I’ve never even questioned before.

A few examples: whenever an ambulance with activated siren passes me by, I have to cover my ears. Being within 20m of such a siren is extremely distressing for me, and would be genuinely unbearable if they didn’t always pass by very quickly. I get that they have to be loud especially so they reach people blasting music in their car, but it still always seemed excessive to me, to the point that I was mad at whoever decided to make them this loud. I only finally noticed about 2 months ago that I have never in my entire life seen anyone else cover their ears when an ambulance passes by. I now assume that no one else finds this nearly as uncomfortable as me, and it makes these observations a lot more understandable.

Similarly, I have a few friends who have a tendency to put on music when we are sitting together and having a chat at home. I’ve always found this very annoying: at best, the music is quiet enough I don’t notice it (at which point you might as well just leave it off) or it is actively making it harder for me to concentrate on the conversation. This is the sort of thing that hits the sweet spot where “it’s not big enough to mention it” and “it is actively bothering me” overlap perfectly. And I’ve never understood it: how can you enjoy putting on music while talking to others when it’s this annoying and distracting? I now assume that, again, I am the only one who has this problem, so it makes sense no one else ever seemed bothered by this.

There are several comparable situations I’ve encountered more or less frequently throughout my life. They have always bothered me, but I didn’t think much of them. It was only through my stay in Oxford that I went through several situations that made it glaringly obvious that this is a me-issue, that I have problems that most of the people around me don’t seem to have. There was one particular event that made me realize that this is not normal: towards the start of my first term in Oxford, I went to a pub quiz with friends. We were sitting together around our table in the college bar, heatedly discussing the questions we were given, yelling over the noise in this cramped space full of people. For some reasonm the noise itself didn’t bother me much that day, but I had an extremely hard time following the discussions we were having, despite me sitting in a rather central position at the table. But the part that drove it home was when halfway through the game, the two sitting on my left and right we’re having a conversation with each other. They were literally talking right through me, each yelling almost straight into my ear. I didn’t understand a single word they were saying. That’s the point where I finally realized that this is definitely not normal.

Since then, I’ve had a couple more situations that illustrated this rather well, and I paid much closer attention to how noise affects me. As a result, I’ve identified a few patterns and feel like I know how to deal with this better now, but a lot of it seems quite arbitrary. For example, although I usually find it uncomfortable to be in loud environments, I’ve experienced being at parties where music was played very loudly, to the point my hearing was noticably worse after leaving, but I was totally fine being there. Generally, I seem to have much less of a problem with loud music, at least when it’s music I like. I’ve noticed this especially at karaoke parties, where I can keep going for hours with the music at a volume that would normally drive me insane.

At the same time, it’s not only about the noise level and whether it is “musical” sound. Certain sounds as well as certain situations are particularly difficult. Hearing babies crying is probably the worst thing I can imagine. Every time I’m riding a bus and a mother with her young, crying child enters, I find myself grinding my teeth and trying really hard not to yell at someone as I’m overcome by this urge to get off the bus right now.

This brings me to the next point: noise affects me in mainly two ways, the first one being that it sometimes makes me very aggressive. This is… impractical, because it makes it difficult to do anything about the noise even in situations where I could technically affect it. E.g. when I’m talking to friends and they put on music, I’m often not able to tell them calmly that this is bothering me and I would appreciate it if they turned the music off or down.

The other consequence is that it effectively interrupts my thoughts. This is difficult to descibe, but that’s sort of what happened in the pub quiz example I described earlier: the sound of my friends talking was obviously entering my ears, but I was completely unable to process any of it.

It can go even further though, as I’ve experienced later. I frequently had movie nights with friends and at some point we watched a movie and two of them started discussing the movie quite loudly while we were watching. This wasn’t even extremely loud or anything, they’re just people with loud voices. For some reason though, this was the worst case I’ve ever experienced. I felt the noise bothering me and making it harder to concentrate on the movie, but it kept building up and getting worse and worse until at some point I could barely see anymore. Before, I thought this only affected my hearing and would annoy me, but in this situation I realized that, as it gets worse, it completely undermines my entire perception and thought process. I was looking at the TV screen and I couldn’t even tell what I was seeing anymore. It still kept going until I the only clear thought left in my head was I need to get out of here. And that’s what I did, I got up and left the room without saying a word. I sat down in the study room to calm down and spent 5 minutes just breathing heavily. One of my fellow studyroom regulars was still in the room studying, and after a few minutes she asked me if I was alright, so I guess I must have looked quite not-well. I noticed that I got a message from one of my friends who had guessed what was going on, and I didn’t even have the energy to reply to them. After 10 minutes I started feeling normal again and returned to the movie a bit later.

This was a bit scary, because it felt like I didn’t have control in this situation and because it was completely unpredictable. If you’d described the situation to me, I wouldn’t have guessed that this could “trigger” anything, let alone the worst experience I’ve had with this so far. On a day to day basis, it isn’t nearly this bad, though still annoying. Perhaps the most common occurrence is that I go out for food or similar with friends or family, and we end up at a place that is so loud I become unable to have a conversation with anyone, in the worst case even those sitting right next to me.

Noticing all of this has at least made me more aware of what issues I have, which makes it easier to recognize when a situation is not good for me, or what activities I should avoid. But perhaps the most important realization was that other people don’t have this problem. In the past, many of these experiences went hand in hand with being mad at someone, e.g. for producing the noise that made me uncomfortable. But since I understood that others don’t experience any of this, it’s completely understandable that they’re behaving the way they are. They’re not inconsiderate, they just don’t (and actually can’t) know better, and expecting them to would be unfair. It might not sound like it, but this is progress to me, and I feel better knowing that this is my own problem and I have to act accordingly, rather than blaming other people for it.

I’m completely unsure whether this is related, but I find that I seem to be more sensitive to sound in non-bad ways than most people as well. On the contrary, I seem to be less sensitive to visual impressions. When talking about anime with the other guy, they once said to me “I watch anime. You hear anime.”. This is mostly a comment on my inability to notice a lot of the aesthetic details, presentation, cinematography etc. (not limited to anime of course) that my friend likes to talk about. I can’t usually comment on those, not only because I only care to a limited extent, but because I simply don’t perceive most of them. A more typical measure of this in Western media would, I assume, be how many people seem to recognize actors across different movies and TV shows. I’m almost completely incapable of doing that, though I should probably mention that I couldn’t care less about actors in movies and I barely watched any Western media in the past couple years. But then on the other hand, I do recognize quite a lot of voice actors in anime, and am interested enough to the point that I often look up who voiced who on MAL. Though I rarely remember the actor’s names, and rather their other roles. I guess this is special, simply because I don’t remember every hearing anyone else talk about voice actors anything like actors in Western media.

Similarly, music affects me in ways that, I assume, is different to other people. For the most part this is impossible to describe or measure, so I’ll bring up the one example where I do have concrete numbers: the right music is an absurd boost to my athletic performance. I know that putting on music to get through your workout is a common thing that a lot of people do, but the effects I observe on myself just don’t seem normal to me. In 2022, I had a pretty stable running routine, running 10km, 3 times a week. I don’t usually listen to music while running. One day, about halfway through my run, I decided to put on some music on a whim. I didn’t look at the time again until I finished my run 1.5h later. At that point, I’d run a half-marathon (the only one in my life so far) at the fastest pace I’d achieved in any run that entire year! Now, I get that music helps by distracting you from the exhaustion, but… I tended to run pretty fast already before that, to the point that I couldn’t go faster without stopping from exhaustion before I’d finished my 10km. I physically shouldn’t have been able to run that fast for longer than 20 minutes or something, let alone for this long. And I still felt great afterwards!

Job Search

Whew, we finally got here! This could be its own article, but as I’m summarizing my year in Oxford here, I think it fits well. To be precise, my job search accompanied me from November 2023 to the start of October 2024, of which I spent the majority in Oxford. Let’s get into it!

Deciding on what you want to do after finishing your degree as a computer scientist, is difficult. The field is huge and can be essentially combined with absolutely anything you could possibly think of. Many people from my department, and even Oxford in general, echoed the sentiment that they had no idea what they wanted to do after finishing their degree. This was, in a way, reassuring. Thinking back to when I finished highschool, I feel like almost none of us had a clear idea of where they were going. At best, they had a rough direction, or picked something that was more or less in line with what they were good at in school or had a basic interest in. And honestly, it feels like that worked out decently for most of us. This is in line with how I see myself: I know I’m practically smart and do my work properly. Of course there are things I have more natural talent or interest in than others, but I believe I could have picked almost any reasonable direction, career-wise, and been decently successful. Computer Science vaguely fitted my interests, so I went with that. I don’t regret a thing.

But then I kept studying, and I knew “okay this time I’m probably starting work for real once I’m done”, and as the end of my undergrad drew near, I realized that I was still facing the same question of where I wanted to go after this, as I did after finishing school. The same process repeated itself during my master’s, in a way. At the same time, this isn’t quite true. Out of the myriad possibilities in the world, I at least narrowed it down to “something computer science related” through my undergrad. Similarly, I narrowed it down further to Machine Learning through my choice of courses in the Master’s degree. Not like this is binding: there are enough more “generic” software engineering jobs where I’m sure I could have a great time. But I went to Oxford, I specialized in ML here, and it would have felt like a waste if I didn’t make use of that. Part of it is just sunk cost, if you will. So, also in an effort to narrow down the far too large number of possibilities, I decided I would focus specifically on Machine Learning Engineering positions or similar.

This was hard. It was harder, and took way longer, than expected. I thought, I’ve got a Master’s degree from a global top university with 1 year of specialized experience in a field that is currently booming like crazy, and everyone is looking for ML engineers. Should be easy, right?

Wrong. I’m still trying to figure out how this works at all, but in any case, it turns out that almost everyone in the field is looking only for people with years of job experience, or a PhD if you’re lucky. Job postings that don’t list a requirement of at least 3 years experience are almost non-existant in this field. I initially applied to the big names like Google Deepmind, and, over the course of the next months, slowly realized that I was wasting my time. I unfortunately started taking notes only after a few months, but I believe I’ve sent about 100 applications. The vast majority of those didn’t receive a reply at all, or a direct rejection. There were only about 10 applications in total where I even got beyond the very initial stage, i.e. I was invited to an interview, an online test or similar. For more than half of those, I got rejected right after that first stage. For the remaining four, I made it to the very end. In two of those cases, I was rejected after the final round because they had a more experienced candidate. Number 3 I rejected myself because it wasn’t ML so I wasn’t that interested in it, and the conditions were relatively bad. Number 4 is the first one where I got an offer, and I accepted it. More on that later.

So in short, if we don’t count the one where I applied to non-ML stuff, I did not get rejected right away in only about 10% of cases, and accepted in 1%. Mind you, I did not have particularly high standards or expectations: I applied almost worldwide, including in Spain, France, Ireland, UK, Sweden, Norway, Germany, the Netherlands, Switzerland, Italy, Poland, the Czech Republic, Romania, Greece, Singapore, India, Taiwan, China, Japan and the USA. There’s not that many jobs I’ve avoided due to my location preferences, mostly those in the Middle East as well as non-remote jobs in the USA, because I really don’t want to go to that country. As a result, I would say I applied to the vast majority of jobs that vaguely fit what I wanted to do, though I started at some point to exclude those that asked for work experience, because I’d noticed by now that those don’t work out at all. In fact, my standards were so low that I would have taken a job that offered less freedom and less than half the salary of what I ended up getting. But they rejected me, heh.

So, what’s there to say about job search? Well, it was interesting. Like I said above, it was much harder than expected, which was sobering and great to keep my arrogance in check. It made me realize that a lot of companies (including many big names you’d recognize) have utterly broken recruitment processes. Before, I’d thought of the technical parts of interviews as where they are challenging me and it’s solely on me to make a good impression, but I realized over the past year that the kinds of exercises they pose tell me a lot about the company. There were a lot of mundane exercises, such as personality tests and classic, automated coding tests. Let’s talk about the latter first.

I believe those automated coding tests are, for the most part, quite useless. Because it would be too much effort to design a different test for every different framework or library you’d expect your candidates to know, they tend to be rather open, in the sense that “as long as you can code, you’re good”. This goes so far that you can typically even code in your programming language of choice, offering 10 or 20 or so to choose from. As a result, you can only test general, competitive programming-esque skills. These typically give you an “in-universe” explanation (e.g. “you’re trying to route trains through a rail-network in a way that minimizes detours” or the likes) that you have to extract the task-relevant information from, then write an efficient algorithm that solves the problem. Now, for 90% of the way, this is a sound start: it tests your ability to extract relevant details from real-world tasks, turn them into simplified, abstract models and then come up with a strategy to solve the assignment in this model.

That said, several problems quickly become apparent, mostly hidden in the word “efficient”. I’ve done a bunch of these coding tests, and almost all of them have very high expectations for an “efficient” algorithm. It wasn’t uncommon for me to spend 10 minutes writing a decently efficient algorithm that solves the problem - but gets rejected because it doesn’t solve that one ultra-strict performance test case, only for me to spend the remaining 50 minutes thinking about a more efficient variant. In fact, these problems tend to be so hard that I timed out on almost every coding test I had to do.

So, this is difficult to criticize effectively because the question at the root of this is, what exactly should you realistically expect from people in terms of algorithmic knowledge/competitive programming skills? Again, I’m completely onboard with this on a fundamental level: optimization is important, and for years we’ve been suffering from shitty software getting slower faster than computers get faster. And I personally like optimization, and think I’m decently good at it! The problem is that, for one, these tests are too strict. They pose problems that are not too hard to solve, but quite difficult to solve efficiently. If you are looking for people with peak competitive programming ability, that might be what you want, but this usually feels more like Executive A bought the “tech recruitment pipeline B” package from recruitment solutions provider XYZ, and that happened to include these tests, even when they’re almost completely irrelevant for the job.

On the one hand, you can get incredibly far with very basic knowledge around algorithms and complexity. I would guess that for 98% of engineering positions, it is enough to know binary search, $\log(n)$ insertion/deletion data structures, understand that sorting can be done in $O(n\cdot \log(n))$, and understand how to derive the complexity of a simple algorithm. Furthermore, actual optimization in practice usually requires almost completely different knowledge. In most applications, you don’t write performance-relevant algorithms from scratch. Usually, you just stick together existing tools. What you then need is good knowledge of these tools to understand where bottlenecks are, be familiar with benchmarking and profiling techniques to find what’s taking up your processing time, and possibly a few things about parallel programming to speed things up further.

On the other hand, competitive programming skills are an extremely niche thing. It may come in handy from time to time, but only in certain roles and even then only rarely. So rarely in fact, that I’ve never seen anyone recommend you practice competitive programming for anything but fun, and coding interviews… And if your recruiting process depends on applicants having spent an inordinate amount of time on building a skill that is likely irrelevant to the role, then you’re not just taking shots in the dark, but your process is becoming downright counterproductive instead.

I’ve gone through a few good recruiting processes, and they all had several things in common. First, they test you on your coding ability in a live interview. This allows for much more valuable insights for them, as they can understand how you think and approach problems, and at the same time it prevents simple misunderstandings or minor problems from becoming roadbloacks that undermine an applicants entire interview performance. Secondly, they test you in meaningful and creative ways that have a close connection to what you’ll be doing at the job, with creative, multi-faceted and obviously hand-crafted exercises. This has several benefits: they can be sure that you have decent knowledge for the exact role you’re applying for, and at the same time you can’t cheese those by having practiced something unrelated. But it is useful to me as an applicant too, because these kinds of exercises provide me with a window into the day-to-day work at the company, helping me ascertain whether I might like it there, and sometimes, they’re just fun exercises to solve and perhaps even teach me something new.

On the other hand, the factory-made exercises at best tell me nothing about the company, and at worst only allow the conclusion that they’re bad at recruiting and don’t care who they’re getting. But it gets worse, and this is where I’d like to get into a few examples of recruiting fails, where the recruitment experience made me genuinely like the company less, and made me reconsider whether I want to work for them.

Example #1

I’m not sure if it’s a bad idea to put company names up here, so I’ll leave them out for now. Example #1 is from a huge Japanese company that you’ve almost certainly heard of. The “coding test” here was especially offputting: it was effectively a Python trivia quiz, with questions such as “What does the expression 15 and 4 return?”. The answer is that, when anding non-boolean expressions, the last one is returned, but only if neither of them is None (or rather, Falsey, which I’m not getting into here). This is so obscure I don’t think I’ve ever seen it in real code, and many people see it as bad style because it’s rarely useful and almost no one knows this syntax. I could have figured out the answer for every single question on this test by opening a Python interpreter and typing in a simple example within 5 seconds. And I would have needed to, because they asked exclusively about obscure syntax details and similar stuff like the example above. This means the entire test had zero relevance to the job. But to top it all off, the entire test was about Python 2. PYTHON TWO! This version was initially planned to be EOL’d 10 years ago. Mind you, I was applying for a machine learning position, a field where the entire software stack in common use is less than 10 years old. You can’t do ML in Python2 at all unless you’re planning to almost literally reinvent the wheel to tie yourself to a long outdated platform. This entire episode is so stupid, I put it on my mental “software engineer jokes” list. All in all, this made me consider redacting my application.

That company also gets bonus points for an utterly broken in-house application management system, which makes the ID of the opening you’re applying to a part of your username, effectively forcing you to create a new account for every position you’re applying to. This is particularly funny because they explicitly tell you to reuse the same account, probably so they can tell if you’ve applied again before the 6 months of break between applications, that you’re supposed to wait, are over.

Oh, since I’ve mentioned account creation, let me just say how much I hate My Workday Jobs. This seems to be another one of those recruitment solutions providers that roughly half the industry is using. What’s so frustrating is that they always require you to create an account. This is already annoying in general, I probably have like 30 completely useless accounts lying around because I applied to some company somewhere. What makes My Workday Jobs particularly bad though, is that every company seems to run their own instance of this platform, so I have to create an account for every single one of them, but then they all ask the exact same extensive set of default questions that takes me like an hour to fill out. I just checked my password manager and noticed I have 7 accounts for My Workday Jobs alone. I’m looking forward to be in a position where I refuse to apply somewhere because their recruitment process sucks.

Example #2

Anyway, looking at more recruitment fails. This one was for a big company that is leaning much more to the hardware side, and fills an important role in todays tech ecosystem. I hold, well, held, them in high regard. Then I applied there. The position I applied to was notable in the sense that I covered not only all of the “required qualifications”, but also the “bonus qualifications”. This happened extremely rarely during my job search, so I got my hopes up immediately. I applied there one July evening - and got immediately rejected the next morning, mere 12 hours later. To quote their reply: “Unfortunately at this time we will not be able to proceed with your application as there are profiles we feel more closely match the job requirements of the position.”

This was weird in several ways. For one thing, recruitment at most companies tends to be pretty slow. Getting a reply within 12h is incredible. Secondly, the reason given seems completely out of place, because I very obviously fulfill all the job requirements. This allowed only one interpretation: someone thoughtlessly drafted a generic answer that does not only tell me nothing, like they usually do, but tells me something straightup wrong instead as the reason for the rejection. I suspected that they actually have requirements for “years of experience” or similar for the role, and they just didn’t list them on the job posting for whatever reason. Or perhaps they’d already filled the position shortly before they looked at my application, which would explain why they were able to reject it so quickly.

I was a bit frustrated with this outcome, but I was busy with other applications anyway and rejections were, in itself, something I was used to at this point. But only two weeks later, I happened to see another job posting by the same company, with the exact same content. At first, I thought that perhaps the post was still up at that point, essentially refuting my second hypothesis from earlier. But no, the posting was super fresh, only about a day old. So I decided to try my luck again, and applied. The exact same thing happened - I got rejected within 24h, with the exact same text, sent by the exact same person. I also confirmed that they’re a real person working at the company by the way.

Now this clearly disproved my second hypothesis, so I was left with the first one, and I was mad. I know that I have little experience, that I am far from the most attractive candidate, and that there is plenty of reason to reject me. I am not complaining about that. Well, maybe a bit, but that’s not the issue here. What bothers me, is that this company that I was holding in high regard was wasting my time by not communicating their requirements properly, not only once but twice, and then sending me misleading rejection mails. I calmed my anger and tried to forget about this episode.

Until a month later, when I happened to see three similar job postings by the same company in relatively short succession. Now, fortunately, the applications there took little effort, so I applied again, with equally little hope. Not sure if I should have been surprised at this point or not, but the exact same thing happened again with all three applications. Rejected within 24h for the exact same reason, letter by letter, and by the same person. At this point I felt like I had collected enough material: I drafted a complaint mail to their global recruiting address. I told them I wouldn’t be applying to them again after this experience, but asked them to at least list all their requirements clearly on the job posting, so they wouldn’t waste other people’s time the same way they did with me.

I wasn’t expecting to achieve anything by doing this, but I had at least a sliver of hope that they might consider being more transparent in the future. To my surprise, I got a reply - from the exact person who’d rejected all my applications! They were offering me a call, to explain why I had been rejected. This was both more accommodating than I’d expected from that company at this point, and also completely missed the point of my complaint. I was mentally done with them and only meant to leave feedback for them to maybe improve their processes. But I was also curious, so I agreed. TL;DR, the recruiter told me they’d rejected me because they wanted someone with at least 2 or 3 years of experience. It was almost anticlimatic how on-point I’d been with my first hypothesis since the first rejection. “So why don’t you put that on the job postings then?”, I’d asked. “It would have saved both you and me our precious time.” Their answer? “Because we want people of all sorts of backgrounds to apply”. You’ve gotta be fu-

I sighed frustratedly. So they’d been intentionally, with full awareness, wasting both mine and their own time then. Great. I was ready to end the call there since I’d found out what interested me, but then they made an interesting proposition: “I’ve found a few other positions that match your profile well, so if you’re interested, I’ll send them to you.” I had relatively little interest in the company at this point, but was still curious where this would take me, so I said yes. Shortly after the call, they mailed me two positions, and I applied to both. They were not quite related to those I’d gotten rejected for before, so not sure why they even sent me those, but anyway: this is the point where this all turns into a comedy. The day after applying to these positions, I got rejected with the following reason: “Unfortunately at this time we will not be able to proceed with your application as there are profiles we feel more closely match the job requirements of the position.” The rejections came from the same recruiter who had personally recommended these positions to me only a day earlier. This was so funny, I laughed out loud when I saw those two rejections one morning. I’m glad I stuck around to see the conclusion of this journey, lol.

Example #3

Oh yea, since we were talking about funny rejections already, I applied to a well-known German startup twice and got rejected both times. 4 months after I applied there. At that point I’d long put them on the “no reaction” part of my list, and the rejection felt more like they were making fun of me to be honest.

Example #4

This is one of the first roles I applied for, namely the one that I was not that interested in because it wasn’t ML. I applied to this almost on a whim, because it was a well known company in the open source software space, which I’d love to support, and they listed Tokyo as the location, which seemed exciting. This process consisted of steps that were mostly fine on their own, but the sheer number and impractical order of these steps made the overall process pretty bad… allow me to go through this chronologically.

I applied to the role in November 2023. I was immediately sent a “written interview” to fill out within the next two weeks. First of all, this was long. They asked a bunch of questions about various aspects of my personality, how I work in a team, my greatest achievements, what I consider important to write efficient or secure software etc. Again, most of these questions seemed fitting and meaningful to talk about, and I didn’t struggle to come up with answers here at all. The problem was the sheer number: I had 40 questions to go through, and ended up typing about 15 A4 pages of answers. If I’d applied any earlier during the term, I would have been too busy to finish this before the deadline.

Next, I was invited to an online personality test. I don’t remember exactly what I had to do here, but it was the usual stuff. I genuinely can’t judge whether these tests are meaningful. Next, I had my first interview with a real person - someone from HR, who asked me a bunch of questions about my background, ambitions etc., all the “soft” stuff. Next, I had to give them times where I was available for technical interviews. What I didn’t expect was that they were going to book not only one, but three technical interviews, on consecutive days. I had a technical interview on Monday, Tuesday, Wednesday morning. It was mid January at that point by the way. When the guy for the Tuesday interview didn’t show up, it was rescheduled to Thursday morning instead. The interviews themselves were all fine by the way, genuinely nice engineers, who asked me all sorts of technical questions. I learned a bunch of new things here, that’s how good interviews should go.

After that, I had to do another (different) personality test for some reason, and then I finally met the recruiter who’d been guiding me through this process so far. It was Feburary by now. We finally got the chance to talk a bit about salary expectations, location etc. We had another meeting a week later to confirm things, but this interview already told me what I needed to know: among other things, the location “Tokyo” was false advertising and I have no idea why they posted it like that. They’re a completely remote company that is not even based in Japan, and do not sponsor visas. Also, the salary wasn’t that good. I rejected the offer I got mid Feburary. After 3 months of selection procedures, including one written interview, two online tests and 5 in-person interviews, the result was that I could have spared myself the entire process, had they just put the salary range on the job posting and not lied about the location.

To this day, I’m wondering why companies so often refuse to put salary ranges on their postings. It’s incredibly annoying because it’s extra work. It also makes me less likely to apply there and it makes them seem less trustworthy.

Positive examples

I’m going to list a few positive examples of selection procedures and other positive surprises here for comparison, which were unfortunately quite rare.

One process consisted of four applied ML exercises, mostly about reproducing experiments and extending them in interesting ways. These seemed genuinely interesting, being about writing a genetic algorithm and then training it via reinforcement learning to beat a game like Pong, or training a generative model to come up with completely new Kanji for modern words, such as AI or airplane. These exercises seemed quite interesting in principal, but unfortunately were completely overwhelming for me, after just starting my second term in Oxford and trying to find a thesis project and supervisor at the same time. I probably spent about 40 hours on this over 4 weeks, and then eventually gave up. I was still working on the first exercise, getting lost in the details of the absolutely terrible framework they wanted us to use for this exercise. I’ve never seen something that encourages writing straight up bad, illegible code like this before. Fighting that framework rather than working on the actual exercise was quite frustrating and I eventually accepted that I don’t have the time to work on something of this magnitude… the kind of devotion that some companies expect already from applicants is kinda insane.

I had another one that gave me a bunch of exercises to solve using PyTorch, that covered a wide variety of things (basic training & optimization of architectures, writing your own backpropagation implementation, understanding and manipulating five-dimensional tensors etc.) that were great in the sense that they were engaging exercises, not too long, neither basic nor super difficult etc. To top it all off, I was invited to a discussion of my solutions afterwards, where I could explain my solutions and was asked follow-up questions. This is a 10/10 of how a good selection process should look like. For the final stage of the process with that company, I was given the task to present a recent (<6 months) paper to their ML engineering team. This was again interesting because it was unlike any task I’d had to do before, and it allowed me to engage directly with their engineering team, which was nice. It was, however, a lot of work to prepare, especially since most of the papers I read weren’t quite that recent.

Another advantage of being in Oxford was, that I was physically close to interesting events. For one thing, I was able attend a sort of recruiting event for one company that I ended up applying for after that, where I got to the final round. I got rejected there unfortunately since they had (as usual) a more experienced candidate lined up, but the selection process itself was pretty good. The practical exercises were mostly logic puzzles and similar. The recruiter was extremely helpful and even set up a feedback meeting with me afterwards, where they told me exactly how I’d performed during the different steps of the process, what I could improve and where I did well etc. This was by far the best feedback I’d received and I was very thankful for it.

Another time, I was messaged by a recruiter looking for people to join their consulting agency’s new office in Oxford. He set me up with the local colleagues and I ended up visiting them at their office, chatting with them and even going for lunch together. That was really nice and a great first impression. If I had any interest in working in consulting, that probably would have convinced me to apply there.

There might be more to talk about, but I had to sign NDAs for some of my selection processes, so I can’t talk about some of the details here.

Learnings

So, let’s get to perhaps the most relevant part: what did I learn from this experience that I can share with you?

Perhaps one of the most important takeaways, at least if you’re very early-career like me, is to look for jobs that are not listed on LinkedIn. Find other platforms, research interesting companies and go to their websites directly, etc. Every job that is posted to LinkedIn gets like 100 applications, and if you don’t have any related work experience yet, you have no chance whatsoever. In fact, I probably did about half of my applications through LinkedIn, and I’m not sure I even got to the first stage in even a single one of those.

As already mentioned, experience is extremely important. This might depend on your field of course, but especially in ML, almost no one will even consider you if you don’t have relevant work experience already. Keep this in mind and accept that you will have to lower your standards and just get what you can to build experience, or maybe aim for easier-to-get internships at first. Keep this in mind when planning your career in general: I did a Master’s, but you may be better off using that time to do internships already. Degrees seem to have little relevance as a bonus, but are mostly used as a fixed condition. I.e. there are many jobs that explicitly expect you to have a Master’s degree in the first place, but for those that do not, you’re better off putting that time into getting work experience instead.

Next, start applying early. Like I said, it took me 11 months to find a job. Don’t expect to find something quicky, and even if you apply to the right position on your first try, then the selection process is probably still going to take 2 months until you get your offer. Plan accordingly. Another advantage of applying early and often is that you get to practice the entire process a lot. I can tell that I’ve gotten better at this, and where it took me several hours to finish a single application before, I can now usually do it in less than one, and often even better than before. Writing motivational/cover letters especially is something that I got much more comfortable with. It’s still probably the most annoying part of the process, but also where I felt the most progress.

Finally, a piece of more personal/mental health advice: keep your hopes in check and don’t ever try to picture yourself in that role or at that company, unless you’re asked for exactly that in an interview question. This might be difficult, and of course it’s easier to e.g. write a good motivational letter when you’re excited about the company, but from my experience, even if I think I have “a decent chance”, I’ll get rejected in >70% of cases. It is incredibly easy to end up imagining your life at that company, and that’s not a bad thing in general, but it is incredibly unhealthy when you do get rejected. I made that experience several times, and even once I was fully aware of this, I wasn’t able to stop those thoughts completely.

The state of the industry and functional workplaces

I have relatively little experience in the industry so far, mainly based on two internships I did in Germany (one at a medium-sized startup, one at a huge German tech company) and then some insights I got through the job hunt.

I started reading a lot of technical blogs this year, among them Nikhil Suresh’s blog. The guy is a data scientist in Australia with 10 or so years of experience, and mostly writes aggressive but well thought out rants about the state of the industry. Trying to summarize his position is difficult, but I’ll try to give a rough overview of his main ideas anyway.

He believes the average developer is extremely incompetent, and just knowing the most basic git commands and writing tests would put you in the top 5% or so from his experience. In order to cover up their lack of competence, they “cosplay at engineering”, pretending that they have skills they don’t and essentially dabble away all day working on projects where their lack of skills has no impact because they won’t ever be used anyway. This is reflected in a symmetry with management which acts similarly, in the sense that the work anyone does is effectively irrelevant, most of the stuff they work on will end in a failed product anyway, but the industry is so dysfunctional that the cash keeps rolling and everyone is pretending to do a good job until the project is dropped for some reason or another.

I’m sure I forgot a bunch of things here, but in any case, he paints an extremely depressing image of the software industry. I found this incredibly interesting, because on the one hand, his articles are full of examples of absolutely insane cases of incompetent and irresponsible people and incidents, that would be funny if you didn’t know that people may have straight up died because of those things. On the other hand, I find that almost none of this resonates with my personal experience. During my first internship, I already had some development experience, but I still learned a lot during that time, felt incredibly productive, and was surrounded by great colleagues who did a great job in their respective areas, teaching me many things along the way. This might be even more the case in my second internship, where I was surrounded by colleagues which were incredibly skilled when it came to understanding C++ compilers, optimization, algorithms and mathematics etc. Similarly, my job search supported these findings. While there is of course a decent chance that I got rejected for most positions I applied to because I didn’t have the word “Cloud” plastered all over my CV or something, almost all of those where I went through interviews and tests etc. showed without a doubt that their engineers know their stuff and are, without exception, more skilled than I am.

While my experience so far is limited, everything I’ve seen suggests that he is wrong. I don’t doubt for a second that he is telling the truth, but I’m still wondering how our experiences could differ so much. This is not a one-off thing: The guy wrote an extensive article about how he found like two tech companies in the entirety of Australia that seem to work like… I don’t know, sane places? I find this incredibly fascinating, because it seems like we are living in completely different worlds, and I cannot for the life of me figure out why our experiences are so completely incompatible.

However, there is one thing that I skipped over here. Remember how I mentioend that I didn’t avoid many jobs due to location preferences? Well, there’s an entire field I avoided almost completely. When I started looking for jobs, I set my status on LinkedIn accordingly, hoping that interesting offers might come my way. Indeed, I was contacted by recruiters on an almost weekly basis. But only by two kinds of companies: tech consulting agencies and automatic trading companies. I rejected all of these immediately, in the case of consulting, because there aren’t many opportunities to work on ML stuff in that space, and in the case of automatic trading, because it seems like it would make me miserable. The latter is difficult to describe, but often their messages were written in this specific way only tech-business bros can produce, that makes you cringe from the vibes alone. I could tell immediately that I probably wouldn’t just not do anything useful, but rather actively harm our society by working there, likely in an absolutely insufferable bubble of… idk, if you’ve seen it, you probably know what I’m talking about.

In any case, I can sort of imagine those spaces being like what Nikhil Suresh describes, and the fact that they seem to be the only ones recruiting desperately enough to message me on LinkedIn, might even support this. It still doesn’t quite resolve my confusion about this whole issue, but I thought it’s interesting nonetheless. In any case, I’m very curious to see if working for a few years as a proper FTE changes my perception of this.

Onward

Well, I’ve already hinted at it, but I completed my job search. So what’s next?

To my great surprise, I got accepted by a Japanese company, and will be moving to Japan in the very near future! I will be working as a Machine Learning Engineer at a startup based in Tokyo.

All of this was a bit surprising: when I applied to the job advert (which was in both English and Japanese), I was told the first interview with the CEO would be in Japanese. At that point I thought, “well, guess no chance then, but I’ll try for fun anyway”. Surprisingly, I didn’t struggle much during the interview, and when I mentioned that my Japanese ability was perhaps my greatest concern, the CEO told me that my Japanese was great and he sees no problem there. Even if I didn’t get the job, this was amazing feedback. I went through another two interviews in Japanese after that, and eventually got accepted.

It’s funny how things sometimes turn out. I never really planned to end up in Japan to be honest. Obviously, I applied to positions there but I didn’t think it would work out, especially not for a position where I was expected to speak mainly Japanese.

In any case, I have by now been working for them part time remotely as I’m waiting for my visa. It should be done soon and I’ll be moving before the end of January. I’m incredibly excited! Expect articles about life in Japan, Japanese work culture and the immigration and apartment hunting process there, when/if I get the time, lol.

Whew

This was way too long. As you can tell by the dates, it took me more than 4 months to write this article from beginning to end, and there’s honestly no excuse for its length. I’ve experienced a ton of things in Oxford and on the side during my time there and I wanted to talk about, or just preserve, so much of it. But I also really need to make an effort to be more concise in the future, or at least split my ramblings into more focused, digestable pieces. Thanks for reading in any case, and I hope it was worth your time!

It was an extremely dense, stressful, exciting, instructive and shaping year for me, and the next adventure lies only weeks ahead. See you at the next article!