Autoexperiment Skill

Run autonomous time-budget experiment loops. Each iteration modifies train.py, trains for a fixed wall-clock budget, evaluates, records in results.tsv, and repeats.

Setup

1. Ensure results.tsv exists with a baseline (exp000) before iterating
2. Create EXPERIMENT.md with your goal, baseline, hypothesis, and constraints
3. Run: /mlx:autoexperiment path/to/train.py

Protocol

Before each iteration

1. Read EXPERIMENT.md for the current hypothesis
2. Read results.tsv for experiment history
3. Identify one change to make (ONE variable only)