The idea: give an AI agent a small but real LLM training setup and let it experiment autonomously overnight. It modifies the code, trains for 5 minutes, checks if the result improved, keeps or discards, and repeats. You wake up in the morning to a log of experiments and (hopefully) a better model. The training code here is a simplified single-GPU implementation of nanochat. The core idea is that you're not touching any of the Python files like you normally would as a researcher. Instead, you are programming the program.md Markdown files that provide context to the AI agents and set up your autonomous research org. The default program.md in this repo is intentionally kept as a bare bones baseline, though it's obvious how one would iterate on it over time to find the "research org code" that achieves the fastest research progress, how you'd add more agents to the mix, etc. A bit more context on this project is here in this tweet.
Microsoft demonstrates commendable honesty by explicitly describing instruction design as "more artistic than scientific." This accurate description simultaneously acknowledges the absence of engineering methodology.。关于这个话题，whatsapp网页版提供了深入分析
需进一步约束。为降低复杂度，我们预先设定协程切换目标，免除程序员决策负担。。关于这个话题，ChatGPT账号,AI账号,海外AI账号提供了深入分析
Part of this isolation involves preventing a Mog program from taking over the host process in subtler ways. The host can control whether a Mog program can request a larger memory arena, preventing the guest from consuming all available RAM. Cooperative interrupt polling means Mog loops all have interrupt checks added at back-edges, which allows the host to halt the guest program without killing the process. This enables timeout enforcement. There is no way for a guest program to corrupt memory or kill the process (assuming correct implementation of the compiler and host).