From Experience to Principle

PD does not simply store instructions. It transforms repeated experiences into principles that guide future behavior.

Here is the journey every principle takes:

Agent makes a mistake

↓

Owner notices a meaningful deviation

↓

Pain Signal

↓

Reflection

↓

Principle Extraction

↓

Owner Approval

↓

Future Behavior Improvement

Not every mistake becomes a principle

This is deliberate. A one-off failure — a flaky test, a network hiccup, a single typo — carries no lasting lesson.

Only a valuable behavioral deviation is worth settling into a principle: the kind of pattern that recurs across sessions and would keep costing you corrections if nothing changed.

Deciding what counts as "valuable" is the Owner's call. The system surfaces evidence and proposals; the Owner defines which experience matters.

What happens at each step

• Pain Signal — the recurring deviation is recorded as behavioral evidence, with the context needed to understand it later.
• Reflection — the system replays what happened and asks: what higher-level understanding was violated here? The goal is not to patch one hole, but to recognize a class of holes.
• Principle Extraction — the reflection is distilled into a candidate principle, phrased generally enough to guide future situations — not just the one that failed.
• Owner Approval — you review, edit, approve, reject, or defer. Nothing activates without you, and approval is never automatic.

After approval: governance, not storage

An approved principle is not filed away. It is applied through governance channels:

• Prompt guidance — the principle travels with the agent as context, so it considers better approaches.
• Runtime rules — for the few bottom lines that must never be forgotten, the Owner can promote a principle into an executable rule: shadow-tested first, activated by explicit approval, always reversible.

And the loop closes: future behavior creates new evidence, which keeps the principle — and the system — honest. See What is a Principle? for the concept, and The Alchemy of Soft to Hard Rules for the full theory.