# Founder's decision journal

Ask a forecast with a date. Ambient Precision asks "Did it happen?" when the date comes, then two thumbs: how good the decision was given what you knew, and how much was luck.

```json
{ "state": { "decision": "Hire a second engineer now instead of in Q3." },
  "questions": {
    "works": { "type": "likelihood", "instructions": "Will `decision` have paid off by 1 March?" }
  } }
```

From the assistant (Claude, ChatGPT), the same journal runs through the connector: `ask_thumb` with `kind: "forecast"`, `resolves_by` and `resolves_when`. After about 30 checked forecasts, the calibration on [/me](/me) starts to mean something.