Proof-of-work is an evidence bank, not a gallery of claims
An evidence bank collects small, checkable signals: a task contract, diff, command output, test result, review note, or short decision record.
Each entry should answer four questions: what was needed, what changed, how it was checked, and what the evidence does not prove.
The shape of one evidence entry
| Field | What it records | Boundary example |
|---|---|---|
| Claim | One concrete statement | Not “strong engineer,” but “added bounded verification” |
| Source | File, diff, command, or review note | A path and command output when actually available |
| Signal | An observable result | The check passed or the result is Unknown |
| Boundary | What cannot be inferred | It does not prove production readiness or external evaluation |
Selecting evidence
- Prefer different signal types over five similar screenshots.
- Keep traceability from the task to the result.
- Do not edit evidence to hide a failed check or an unknown result.
Canonical lesson source · JavaRush
Open JavaRush material