It’s insane how many reshares of this post emphasises more on “while sleeping / before bed”. I feel the issue is most people can’t wrap around the big ideas that are coming out. And try to hype things simplified into their life like working while sleeping or just writing AI slope
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.
https://t.co/YCvOwwjOzF
Part code, part sci-fi, and a pinch of psychosis :)
@OpenAIDevs Could you please add a feature in Codex app where if I click on a file name from the output, instead of opening in another IDE, it will open a side tab like current "diff panel" and view the file inside the app.
But now I think it's more about "learning how and what to learn" rather than spending time to systematically learn stuffs. Our universities were inherently built to teach some topics, so the paradigm is shifting to fast, we need to figure out the purpose faster.
Current education system won't be able to serve the AI generation that's coming. I understand it's kind of unfair from teacher's point of view to finish the assignments with AI. But if AI can solve an assignment, what are we learning from that assignment?
But now that lots of the "defined works" are done by AI, we need to define the work again now. I think the bottleneck currently is imagination of how things will shape in coming 1-2 years.
Coming back to university education, yes the point is still about learning how stuffs work. But the cost of learning has reduced drastically, both in terms of time and effort. Earlier you still had to search for materials to learn, before that it was books.
I still do struggle calling and feeling AI generated codes my own. All the years codes have been a small corner of personal creation. But now that I am instructing someone(*thing) to build the code, it feels not so personal.
I’ve been working on something for the beginners… and it’s finally time 👇
This Thursday I’m dropping a FREE trading eBook — the exact foundation, mindset, and simple frameworks I wish someone handed me on day one.
If you’re trying to get out of the “confused beginner” stage and actually understand what you’re doing… this is for you.
Like and Retweet if you want it the moment it drops.
Time to level up 🔥
@FPLMate Cause there were 4th highest number of Wildcards played in the history. So as you know, price changes are not as predictable during the WC window.
@FPLMate I think could you take a moment to explain before each section what to expect from it.
Also explain what to expect in the whole video at the beginning.
Another suggestion is do both Buy and Sell of each type back to back. For example: Sell DEF, Buy DEF, Sell MID, Buy MID etc.
Hey @BrawlStars , many a time nowadays I don’t see the buttons. That way I can’t do anything in the game, most importantly I can’t play. It’s frustrating.