GitHub radar
Kadath is an evolutionary runtime for creating superior AI agents
KADATH iteratively evolves a population of AI agents towards a solution to your task.
KADATH is an open-source multi-agent evolutionary framework built on top of Hugging Face smolagents by one person (i3t4an). This framework uses a bio-inspired methodology to find solutions to problems through competition between agents across multiple generations (epochs). Each epoch is evaluated against a static benchmark, and the Darwinian principle of “survival of the fittest” is applied (keeping the top 30%, letting mid-tier agents mutate based off of the successful mutations of the top agents, and culling the bottom tier). After running for 10 epochs, the maximum fitness went up from 18 to 91, showing that this framework can improve over time without needing to optimize each agent individually.
Why a vibe-coder should care
Instead of fiddling around with prompts to try and get your agent to work, KADATH will search the solution space for the best possible solution. Just give it your problem, how many rounds and agents you want, sit back, and let it do its thing! You’ll need Docker and an OpenAI API key, but if you can execute a shell script, you’re golden!
▌ More finds