Tool Profile

ai-trains-ai

An open-source experiment where an RL agent writes RL training jobs.

At a glance:

First seen:2026-07-14

Last seen:2026-07-14

Sightings:1

Source:github.com

What it is

An open-source experiment where an RL agent writes RL training jobs.

Why developers recommend it

A concrete, reproducible nested-RL prototype with a clear cost and limitation discussion.

Hacker News evidence

2026-07-14

Its author described 1,750 training jobs, hidden-eval feedback, and limits to small agentic tasks on the Prime-RL stack.

Show HN: I RL-trained an agent that trains models with RL (for ~$1.3k)