P5: T3 Evolves Capability¶

T3 evolves capability by learning from examples and experiences — explicitly.

Capability improvement must be tied to specific examples, experiences, and evidence. Release scorecards report which harness improved by how much on what evidence — not a broad "release N is better than N-1."

Why it matters¶

Implicit improvement is unfalsifiable. If you cannot point to the harness, the delta, and the evidence, you cannot tell real progress from noise, and you cannot trace a regression back to its cause.

How it is realized¶

Per-run evidence comes from harness-owned verification (C5). Turning that evidence into measurable, per-harness capability deltas is handled by the evolution-telemetry work and is tracked separately, so this principle is partially validated today. See How it works → Learning from repetition.

Primary surface: release scorecards and per-harness telemetry.