I originally planned to start writing after the lab dinner on Wednesday when the holiday officially began, but I kept putting it off, then figured I might as well wait until after submitting the paper. Today I finally finished tidying up my dorm and have time to write.
Honestly, I don't really know what to write either. I look busy, but I don't know what I'm busy with. Let's take this as a summary of what I've learned before, and make a plan for the learning ahead.
Let me set the overall tone first: since AI is an ability amplifier, we need to spend more time learning and accumulating more ideas. So the main thing ahead is still learning — more detailed and deeper learning. As for internships and work, although I'm anxious about them, I vaguely feel they're not actually that important. These are all external evaluation criteria: joining a company as an intern, and later joining a company for work — in both cases you're making someone else's product, realizing someone else's ideas. So why not try making my own product? Anything is fine, good or bad. Taking things I built myself to look for internships and jobs — I think that helps too.
Full-stack (the plan for now is frontend
Recently I've been learning frontend — reading TypeScript, reading React, and building some projects hands-on. I have some project ideas but haven't started building them yet; that should take quite a bit of time, including reviewing knowledge along the way, and memorizing some material for internships — probably the focus of next semester. The ultimate goal is of course full-stack; one step at a time. Start with web pages, then make an app, and gradually get into backend and databases.
Agent
For agents, I feel what I've learned is still too rough. I want to go deeper — watch more of boj's courses and livestreams, look at frameworks like LangChain, look at SDKs and APIs. Also hooks, skills, and keep following new technologies.
Including the frontend skills released by many seniors.
Memory
Including the subdivisions of memory — I think memory is quite important. My previous understanding was just file systems or RAG; that understanding was too shallow. Actually, because of the internship, I read Professor Bai Ting's survey paper on memory and got in touch with many memory products, and only then did I realize that's not the whole story. There are still things I need to understand more deeply.
Paper
January 29 was the ICML submission deadline. A few days ago, my senior led me in a paper-writing sprint. I drew many figures and tables, and got taught a lot by my senior — how to draw figures, how to make tables, how to typeset and write a paper. I learned a lot. We even pulled an all-nighter together, which was truly unforgettable. Later I need to carefully study my senior's approach to paper writing. Next semester, for graduation, I should publish another paper — but let's set that aside for now; it's not that important for the moment. Related matters can be moved from memory to disk.
Internship
After the 8 a.m. paper deadline, I went to print my internship materials. The next day (the 30th, yesterday), I went to intern at a memory startup for a day. I also got to know a few guys who were all pretty nice. I learned a lot that day, saw many memory products, and some tools those guys were using. But for my own considerations, I don't plan to go next week — I'll focus on my own things. As for summer internships and regular internships, I want to set them all aside for now and build my own products first.
I think internships are all external evaluation criteria. Everyone tells you how important regular internships are, how important summer internships are for conversion to full-time offers. But the essence of the problem is still whether you have the ability to build, the ability to fight real battles. Contributing open source to other people's projects — same thing; let's pause that too, and first try building my own products. There's no time left; start creating. I want to take my own portfolio to apply for jobs and internships. I think my thinking also fits the trend of AI making junior engineers fewer. Of course, the fundamentals must be solid too.
Be Open
I've actually been trying to become more open, trying to meet and talk with more people. A while ago, working with my senior on the paper, and then the brief one-day internship experience — both deepened this belief: engage more with people, without setting any limits.
I'm also trying various ways of dealing with people. Including wanting to join the Kubernetes release team as a shadow — so I need to start preparing now.
I also need to train my expression skills — including using voice input more, like typeless, autotyper, and so on. I wonder if they could also help me get rid of my stutter.
Exercise
This morning I did some squats after getting up. I think it's also important to slowly do some exercise within my ability. My girlfriend often nags in my ear that once a man passes 25, he's basically 65... My senior also strongly emphasizes the importance of exercise — I must take it seriously.
In the end
Nothing much left, I feel. It's nice to do a summary every once in a while. I typed super fast — it's all just rambling.
Recently I've been using Things for todos and some project management — feels pretty good; I hope I can build the habit. Telegram is for saving links to articles I've read.
Later I might do some organizing of Agent memory products.