An agentic harness for local AI models with the ability to build no code RL environments, get a taste of what test time compute and recursive self improvement can look like, and two "new" training methods --> PQLORA & IL ( intuition Learning ).
I wanted to build this to pack a lot of the things I work on inside one single place.
I've been working on RSI for a long time, and I didn't know TTC was an actual thing, but I was working on it, thinking it was "my" idea. When I learned it was an active research field, it motivated me to finish my work, and publish it fully open sourced.