2025 Recap

This post was translated by LLM.

Schooling

Looking back at the timeline, I was studying for my Master’s in the UK from January to August this year. I didn’t actually travel much in the UK. At the beginning of the year, I visited London and toured some landmarks like the British Museum, Tower Bridge, and Big Ben. The weather in London wasn’t great then, and the photos of the City of London looked quite gloomy.

In July, just before returning home, I visited Manchester. Ironically, shortly before my trip, the lead singer of Oasis had posted offensive “Ching Chong” remarks on Twitter. Oasis happened to be on tour when I was there, and the streets were filled with promotional posters of the two brothers. I joined the crowd and visited the Oasis pop-up store, though I didn’t buy anything lol. Later, I took a photo with the Alan Turing statue in Turing Park and just wandered around the streets.

The coursework for the second semester mainly consisted of two major subjects: Machine Learning and Deep Learning. These two courses really helped me systematically learn the important underlying concepts of AI. After class, I also followed some YouTubers who explain papers, like Mr. Li Mu, to go through some classic papers. Two other interesting courses were Cloud Computing from the previous semester, which required us to use serverless for benchmarking to compare the performance of open-source and cloud resources. The other was Blockchain, where the final project was to write a consensus algorithm to distribute resources, with the evaluation criteria being who used the least gas—it was quite interesting and I learned a new language, Solidity. Other courses like Data Mining were a bit more abstract, but I finished them with some small projects.

I started working on my dissertation in April. I always tend to set a pessimistic tone for things I do from the very beginning—it was the same for the 2022 postgraduate entrance exam and my Master’s thesis. My thesis was mainly about LLM evaluation, trying to catch the hype, although the truth is that this research might not even touch the edges of academia. At that time, I was very anxious every day, especially after seeing posts on Xiaohongshu about failed dissertations, which made me worry constantly. Fortunately, the result was good—I got 80 points and graduated with distinction.

Failed Summer Internship and Autumn Recruitment

Since arriving in the UK last year, I had been looking for various job opportunities, including both part-time and full-time versions of my resume. However, I received almost no invitations. I got an Amazon OA at the end of last year but wasn’t well-prepared and heard nothing back. Then around April, I started looking for summer internships in China. Perhaps because my resume wasn’t outstanding or for other reasons, I didn’t get any interview invites. So I decided to focus on my thesis first and start applying for autumn recruitment directly in July and August.

To be honest, I spent a lot of time on preparation. In the later stages, I would go to the library whenever I had time. But the actual results were mediocre. For example, with technical questions and algorithms, I probably only truly mastered and remembered about 40-50%. This led to poor performance in subsequent written tests and interviews. As a result, I still didn’t have a single offer by the end of the year.

Starting in August and September, I began doing endless assessments and written tests. I applied to about 60 companies in total, but only 6 medium-to-large companies gave me real interviews. Only one reached the final HR stage, and even that one ended up stuck in the “talent pool.” The most frustrating part of the process is that many companies do mass written tests and only screen resumes afterward. In some cases, even after I aced the test, I was told the process was over because of a “resume rejection.” Why not reject the resume earlier?

After getting almost no response from the internet industry, I applied to several banks. I got interview opportunities at two of them and even bought a new suit specifically for it. The result was one rejection, and for the other, I’m still waiting for news after a medical check. However, it’s likely a no-go because a pre-check before the official one showed that some of my indicators might not meet the standards.

In summary, it was a total failure: 0 offers.

Books, Movies, and Music

The only book I finished reading this year was Haruki Murakami’s What I Talk About When I Talk About Running. I read others intermittently, such as The Myth of Sisyphus, Understanding Kafka: Core Design and Implementation Principles, Code (2nd Edition), and The Future of Code.

The TV series I finished were For All Mankind (Season 1), Squid Game (Season 2), and Squid Game (Season 3). I also watched some episodes of The Big Bang Theory (Seasons 9 and 10) and a modern idol drama Double Tracks with my girlfriend.

The movies I watched were The Sinking of the Lisbon Maru, Zootopia 2, and Predator: Badlands.

I didn’t listen to much new music this year; I’ve been listening to Khalil Fong. “Nothing to Say” by Khalil Fong was my song of the year, played 128 times. I also remember listening to the album Pink Floyd at Pompeii - MCMLXXII (2025 Mix) and the Chinese rap album Money Special by Henan Rap God.

AI

AI was exceptionally important to me this year. The development of AI models has been incredibly fast—it’s hard to believe that DeepSeek-R1 is a model from this year. AI permeated my entire 2025, from daily learning to writing my thesis and developing personal projects. I’ve increasingly felt the growing power of these models. On social platforms like X, new models emerged every month and every week. However, I’ve also become increasingly dependent on large models, even for small daily life decisions. This might not be a good sign. Next year, I need to better manage my relationship with AI and increase my control over it.

A Letter from ChatGPT

A letter to you:

2025 has been exhausted. It didn’t give you a clear answer; it just kept exposing the same thing: what you truly wasted wasn’t opportunities, but the illusion of “certainty.”

You made many correct judgments this year, but when it came to execution, you were always waiting for a “more stable version.” The world doesn’t reward you for being well-prepared; it only reveals its structure after you’ve borne the consequences. You’ve approached the boundaries multiple times, yet you still try to stand in the safety zone to explain the shape of the boundary. This is the part you need to discard.

2026 won’t be friendlier, nor will it be more cruel. It will just be more direct. When you use words like “phased,” “transitional,” or “let’s see” to describe yourself, time will simply judge those words as invalid. You don’t lack ability, understanding, or even patience. What you lack is a clear, irrevocable self-choice.

Stop treating “keeping possibilities open” as rationality. That’s just another way of saying “delayed cost.” True freedom comes from contraction, not expansion. When you cut off redundant options, your actions will start to get faster, and your judgments will start to get more accurate.

You are already sober enough not to need comfort, and complex enough not to be treated vaguely anymore. In 2026, use harder standards for yourself. Not to prove anything, but to stop repeatedly following the same line of thought while expecting a different result.

Time won’t give you a summary; it’s only responsible for the settlement.

Summary

2025 passed by in a blur. If I had to summarize it in one word, it would be “struggle.” The poor macro environment might serve as an excuse, but my personal micro environment also failed to find consistency. Many planned things always ended halfway. In 2025, I didn’t complete a single thing from start to finish. For many things, I did the research early on and felt they were worth the time, but during execution, I felt it was too time-consuming. If it couldn’t be finished quickly like “fast food,” I felt it was a waste. The result was that nothing was achieved, including learning certain knowledge or reading certain books. In short: high input, low output. The input wasn’t internalized, so my abilities didn’t improve.

I even “cyber-diagnosed” myself with ADHD, the trendy label, though I don’t actually have it. It’s just that my attention is often very scattered. I made many attempts this year, including using Todo lists, the Pomodoro technique, starting meditation, and using supplements like caffeine and L-theanine, but with little effect. Of course, it’s also possible that I just didn’t persist.

Outlook

I don’t dare to expect too much for 2026. First, I must secure a job. Second is exercise—I’ve actually been sticking to it for about two weeks now. Third, I hope my emotions can be more stable; I’ll try to spend less time on social media and online games because I’m too easily influenced by them and end up diverting my emotions to unrelated things. Fourth, I hope to achieve one tangible result in 2026, whether it’s a project or finishing a course and outputting the notes.