CoRT makes AI models recursively think about their responses, generate alternatives, and pick the best one. It's like giving the AI the ability to doubt itself and try again... and again... and again.
YES. I tested it with Mistral 3.1 24B and it went from "meh" to "holy crap", especially for such a small model, at programming tasks.
- AI generates initial response
- AI decides how many "thinking rounds" it needs
- For each round:
- Generates 3 alternative responses
- Evaluates all responses
- Picks the best one
- Final response is the survivor of this AI battle royale
Mistral 3.1 24B non CoRT
Mistral 3.1 24B + CoRT
pip install -r requirements.txt
export OPENROUTER_API_KEY="your-key-here"
python recursive-thinking-ai.py
The magic is in:
- Self-evaluation
- Competitive alternative generation
- Iterative refinement
- Dynamic thinking depth
Found a way to make it even better? PR's welcome!
MIT - Go wild with it
English (US) ·