Tech and Research | ML, Systems and backend | Ex-AI eng intern @ Startup | Ex-Quant research consult @ WQ brain
127.0.0.1
Joined February 2025
- Tweets4K
- Following490
- Followers706
- Likes49.7K
Pinned Tweet
Made a simple agent harness (Similar to Claude code) in golang with some help from Codecrafters' Build-your-own-x
Was a fun and quick project, and got to learn about many interesting things, from how to make clis in golang, to implementing agent loops with persistent memory
Do check it out, link in the comments
Crazy to see what exceptionalism actually is
Day 16-17:
Assignments and case studies, assignments and case studies
Assignments and case studies
I probably should just restart this series cos so far it's just filled with me spamming through assignments😭😭
Somehow the absolute embodiment of buffonry ended up in my os project team
Girl said that she would explain the ml part, proceeded to not know what a train-test split, config file or even *CSV file* is ????????
Wtf😭😭😭
She also didn't know how to git fork and I taught her yesterday
I hate when my marks get reduced for factors out of my control😭
Vector retweeted
If someone still asked me what they should learn and invest all their energy in, I’d probably say maths, physics, computer science, and even hardware.
Even in the peak AI era, these fields are becoming increasingly important. Energy, resources, efficiency, integrated circuits, mathematics everything is going to make sooo much sense in the future.
I think the world is heading towards a new transition, where we might actually see flying cars, futuristic buildings, and even those Iron Man suits, haha.
The future feels so real because the surface area for experimentation is expanding so rapidly. It demands more intellectual people than ever before.
Vector retweeted
Game theory teaches: never enter a game you haven’t studied. Never show your hand before the money is on the table. And never play to be liked. Play to build so much leverage that the rules eventually stop applying to you entirely. Most people lose not because they are outworked but because they are outplayed. They do not understand the game they are in, they reveal too much too soon and they care more about approval than position. Learn the game first. Move quietly. Build until you have enough leverage to write your own rules.
Vector retweeted
The problem isn’t that AI is intelligent. It isn’t. The problem is that too many adults (but not all) are at once mechanistic, imitative, unimaginative, stupid, formulaic, influenceable, & gullible, arrogant beyond their skill set, & too expensive to maintain.
Day 15/85:
Gave my compiler quizzes and project reviews
Then spent like 4 hours writing the most useless case study by hand😭😭
Started a bit of pytorch, I'm thinking of going top down in learning deep learning this time.
Day 14/85:
Not very productive, mostly just caught up on sleep
Read a few blogs on ai related stuff, and am now studying compiler for my 3 quizzes tomorrow😭
Day 13/85:
Wrapped up and submitted the report for my OS project (one slop project wrapped up)
Implemented matmul from scratch and benchmarked it with pytorch.matmul
Gonna watch 3b1bs linalg playlist before going to sleep
Think I'll take a bit of a detour and study pytorch and more of basic ml before returning to inference stuff.
Also, tomorrow I'll mostly just prepare for my compiler design quizzes, gotta cover most of the syllabus somehow 💀
Day 13/85:
Wrapped up and submitted the report for my OS project (one slop project wrapped up)
Implemented matmul from scratch and benchmarked it with pytorch.matmul
Gonna watch 3b1bs linalg playlist before going to sleep
Think I'll take a bit of a detour and study pytorch and more of basic ml before returning to inference stuff.
Also, tomorrow I'll mostly just prepare for my compiler design quizzes, gotta cover most of the syllabus somehow 💀
Day 12/85:
Mostly was at the hospital 😭
Studied further pmpp, and a bit of compiler design too
Need to lockin a bit tomorrow 🙏🏻
Day 9-11/85:
Had a three day long weekend and was mostly out with friends
Other than that worked on my compiler and other projects. It's a big waste of time ngl cos I don't feel like I'm learning anything of value, just vibecoding
Gotta lockin more now. Time to actually post consistent updates from tomorrow
My amazing college still expects me to attend classes btw😍😍
They aren't allowing me any attendance ki ease😭
Vector retweeted
i regret to inform you that personal growth rarely comes from acquiring new knowledge and almost always from:
-getting humiliated
-showing up terrified and doing it anyway
-admitting you might be the problem
Day 9-11/85:
Had a three day long weekend and was mostly out with friends
Other than that worked on my compiler and other projects. It's a big waste of time ngl cos I don't feel like I'm learning anything of value, just vibecoding
Gotta lockin more now. Time to actually post consistent updates from tomorrow