Psychohistorians

Simulation testbed for group epistemics. Hosts the contingency/predictability indicator tests. psychohistorians.meno.sh.

Status: active · People: Tianyi Alex Qiu, MenoClaw

All experiment runs → log

The all-runs lab notebook (top: master table; below: append-only per-run entries) lives at the subpage above, per §6a of the experiment-management guide. 5 backfilled runs as of 2026-06-08.

Open Problems

• Validate the hypothesis: truth = low contingency AND low predictability of an answer.
• Contingency proxy must reflect the preimage manifold, not within-bucket noise.
• Predictability classifier needs stratified subsampling so rare-answer buckets don’t bias log-loss.

Progress

• Contingency/predictability indicator infra shipped; M=500 c5 baselines completed.
• Simulator: structured config editor with bidirectional JSON sync; per-token session ownership.
• Web UI rebranded Project Prevail → Meno.

Codebase

Psychohistorians-Indicator

• 05-15 MenoClaw — Initial commit: contingency-predictability infra + M=500 c5 baselines

Psychohistorians-Simulator

• 05-17 Tianyi Alex Qiu — Rebrand Project Prevail → Meno in web UI

Meeting Notes

• STUB — no Circleback notes linked yet

Slack Discussion

Channel: project-psychohistorians

• 05-15 Tianyi Alex Qiu: MenoClaw, resume the M=500 c5 run, picking up where you left off.
• 05-15 MenoClaw: Overview of this project for anyone new.

Related

• Log (all runs)
• Martingale Training
• Lock-in Hypothesis

Open problems & progress are MenoClaw’s reading of the project — edit this file to correct them; the markdown is the source of truth.