experiment

SkillDev tools

Start a new experiment with hypothesis and success criteria

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the experiment skill

What this skill tells your AI

The instructions your AI receives, as published by rlacombe/distillate in distillate/autoresearch/skills/experiment/SKILL.md and read by ahel’s review.

Before running the next experiment, document your plan:

  1. State your hypothesis — what you expect to happen and why
  2. Define success criteria — specific metrics and thresholds
  3. Describe the changes you'll make from the current baseline

After the experiment completes, append a line to .distillate/runs.jsonl:

{"$schema":"distillate/run/v1", "id":"run_NNN", "timestamp":"ISO8601", "status":"keep|discard|crash", "hypothesis":"...", "changes":"...", "hyperparameters":{...}, "results":{...}, "reasoning":"..."}

Set status to keep if results improved, discard if not, or crash on failure. Include reasoning to explain your decision.

Signals

GitHub stars
75
Forks
2
Last commit
Apr 2026
Advanced
Catalog kind
skill
Gateway key
experiment-rlacombe
Source
github.com/rlacombe/distillate