Experiments Index

Status: Experiment index Audience: Team, AI agents Purpose: Entry point into dated experiment logs, evaluations, retrains, and benchmark changes

Use this folder for dated, narrative experiment notes:

• benchmark changes

• evaluation rounds

• retraining results

• acquisition-wave measurements

• autoresearch observations and decisions

Current logs

• 2026-03-17 — ML detection iteration log

• 2026-03-20 — Multi-instance evaluation and improvement

• 2026-03-21 — Anonymous holdout rework

• 2026-03-21 — Acquisition wave retrain and evaluation

Recommended log shape

Each experiment note should stay compact and follow the same structure:

1. Context

2. Problem or goal

3. Commands run

4. Results

5. Interpretation

6. Decision or next step