Forked @karpathy's autoresearch so it runs on a Windows gaming PC with a desktop RTX GPU: github.com/jsegov/autoresear…
I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then:
- the human iterates on the prompt (.md)
- the AI agent iterates on the training code (.py)
The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc.
github.com/karpathy/autorese…
Part code, part sci-fi, and a pinch of psychosis :)
segov retweeted
Replying to @jonsegov
Computers start eating every other profession and people’s reaction is, don’t learn about computers?
Hot take: CS is going to have an enrollment boom in the 2030s that'll be bigger than the one in the 2010s
Students are fleeing computer science.
US enrollment in computer and information science programs fell more than 12% for undergraduates and more than 10% for graduate students this fall, per preliminary data from the National Student Clearinghouse.
Over two years, undergraduate enrollment is down 18.4% and graduate enrollment 27.1%, reversing years of growth as AI cuts into entry-level coding jobs.
I'm as AI pilled as they come and also a MASSIVE house music fan. Respectfully this is hot garbage and just goes to show that human taste is still important in creative disciplines.
We're living through biblical times
Today, the promises of transhumanism and posthumanism are multiplying. These are ideological currents according to which, sooner or later, technology will eliminate all disease and pain, as well as aging and even death itself. Precisely because scientific progress is a supreme achievement of human intelligence, it must not be confused with a privilege reserved for millionaires seeking immortality, while entire peoples suffer from poverty and oppression. In the face of this mirage, which promotes the illusion of human beings perfected by cybernetics and biotechnology, Christian anthropology reminds us that a limitation is not a manufacturing defect, but rather the very condition of human grandeur. #ApostolicJourney #Peru #MagnificaHumanitas vatican.va/content/leo-xiv/e…
3
118
quick fixes considered harmful: you might be making your agents write worse code when you ask for the smallest or quickest possible change
what seems like a harmless bandaid may lead to that pattern being copied by the next agent. before you know it that pattern has spread everywhere
we used to write bandaid fixes by hand because we rarely had the time to do the right thing. agents change this calculus. sometimes you want it to fix the root cause instead even if it’s more invasive
segov retweeted
“Avoid people who are full of self-pity, and people who feel like victims. Avoid them like the plague.”
- Charlie Munger
The world was built by contrarians that turned out to be right, society doesn't advance by being content with the status quo
being contrarian is good. being contrarian and right is even better.
when i said “non-engineers will build quality software,” people looked at me like i was crazy.
i hear brian at airbnb got the same reaction to “strangers sleeping in your home.”
the best ideas often sound wrong before they become obvious.
1
77
If you're actually still using agents on your machine you absolutely need to take some time to play with cloud agents, they'll completely change your workflow
it feels like cloud agent adoption is where claude code was at this time last year. that’s going to change over the holidays & it’s going to be insane
3
1
137
I'm a certified waterfall hater and I actually kind of agree with this.
Let me me explain: in a traditional waterfall environment there isn't time to expirement and prototype, but in a AI native one there is. Now you can easily create multiple prototypes for how to solve a given problem, take the learnings from each of them, and then plan out how to productionalize it accordingly.
Waterfall felt soul crushing to me because diving head first into a problem and figuring it out along the way is one of my favorite parts of being a SWE. It's also what separates it from other engineering disciplines imo (there's a reason we have hackathons, not planathons). But we're currently at a place where the two can coexist, so waterfall might be aight tbh.
Hear me out:
We hated "waterfall" because after Big Upfront Planning, implementation took FOREVER, and by the time we were done, the plan was out of date
But with AI, implementation is rapid (100x faster, or more, than before)
So Big Upfront Planning can now make sense?
1
1
120




















