Project
Agora
A state-first reality-simulation OS: agent populations evolve under Markov rule dynamics, the system predicts macro-dynamics and evolves rules toward targets, and every state lives as plain files in a vault.
Status Active
Agora
Can a vault-native simulation OS forecast and steer agent-population macro-dynamics under explicit targets, with every world, rule, run, and channel stored as plain, diffable Open Knowledge Format files? Populations evolve under a Markov chain of rules; forecasts and rule evolution are scored against target distributions rather than qualitative judgment. Worlds, rules, runs, and channels live as JSON, JSONL, Markdown, and Canvas in an Obsidian vault so the entire operating state remains inspectable text.
Calibration evidence is public in the evidence pack: a 30-tick opinion-diffusion world ends at adopt 0.600 / undecided 0.388 / reject 0.013; a Markov-plus-Kalman forecast records error 0.4513 against its target; an SIR epidemic world holds susceptible at 0.838 by tick 40. The numbers are small and provisional; they exist so forecasts and rule changes can fail in public.
The brain layer is verifier-first and deliberately powerless: a CompactCTM-mapped shadow predictor forecasts world state and is scored by external Brier and log loss, with no control authority until predictive credibility is earned. Technical reports publish the measured calibration loop; the agora / agora-agent-brain source trees remain private.
Reports
- Report
Agora CompactCTM Shadow — Stage-1 Prediction Without Control
Held-out CompactCTM shadow predictor fails to beat an empirical prior on Agora opinion-diffusion transitions; claim class is non-parity, control authority false, not promoted.
- Report
Agora Calibration Runs
State-first simulation calibration on opinion diffusion and SIR worlds: scored forecasts and rule evolution against explicit targets, not vibes.