Show HN: A local rig to test if AI social simulation predicts reality

Category: devtools

Tags: ai-simulation, llm-evaluation, social-simulation

Score: 6.0/10 (Innovation: 6, Technical: 5, Documentation: 7, Utility: 6)

A local calibration rig that tests whether multi-agent AI social simulations actually predict real-world reactions better than a single LLM call. It provides preliminary evidence that crude swarms do not beat a baseline, shifting the burden of proof onto the simulation category and offering a reusable methodology for honest evaluation.

Target audience: AI researchers, ML engineers, data scientists

Repository: https://github.com/zzvimercm-git/mirofish-calibration · Python · MIT

View on Hacker News