ai-trains-ai - RL agent that autonomously trains AI models
Show HN: I RL-trained an agent that trains models with RL (for –$1.3k)
ai-trains-ai is an open-source project featuring an AI agent trained via Reinforcement Learning to autonomously design and execute training pipelines for other models. Built on the Tinker framework and leveraging Runpod GPUs, this agent writes complete prime-rl jobs, including environments, rewards, and hyperparameters, to teach small models specific tasks. The system demonstrates a meta-learning breakthrough where the agent improves its own training strategies over time, successfully transferring skills to unseen task families. Everything, from the LoRA adapter weights to the orchestration scripts, is fully open-sourced for the community to explore and extend.
The agent started making better models, not just working ones.
- callamdelaney
AI trains AI already, agents are happy to spin up real training pipelines for deep learning or regression models or whatever you want right? I guess the advantage to your project is that it provides a framework to allow the agent to access extra compute?
- bitbasher
Lots of emoji in that readme. Was it mainly codex?
- saberience
Can you explain how it works?
What problems would it do well on and why?
Where would it start to fail/break?
What are the limitations of a system like this?
When you vibe code a system in a complex area like RL, you basically have zero understanding of what its actually doing, whether its actually any good or not, what you're actually benchmarking, and when the system would fail.
It's the blind leading the blind.
- nnevatie
I'm curious to whether the recursively trained models degenerate to troglodytes after a couple of generations.
- azertyvode
How do you prevent the agent from reward-hacking the hidden eval? e.g. writing training data that effectively leaks the eval distribution rather than teaching a general skill?