Skip to main content
In plain terms: This is the heart of Forge. The records why every engineering decision was made, not just what the ship is. It is the event clock from the Two-Clock Problem.

​
What it does

L2.5 stores replayable engineering decisions. Every engineering decision (proposed, drafted, approved, rejected, superseded) is recorded with its full reasoning context:
  • what triggered it
  • which rule version applied
  • what evidence it cited
  • who approved it
  • what it superseded
  • its

​
Append-only, not an audit log

The ledger is append-only: decisions are never edited or deleted. A wrong decision is superseded by a new decision that explicitly references and replaces it, so the full history is always preserved. This is different from an audit log. An audit log records what happened (“user X clicked button Y at time Z”). A decision ledger records the reasoning chain (“decision D was made under rule version R, citing evidence E, by approver A, superseding prior decision D-prior”).

​
Why it is the most important layer

Because every decision records the exact rule version it was made under, the ledger can answer the question that matters most when something changes:
“Which decisions need to be re-evaluated, and in what order?”
That single capability is what turns a rule update from an open-ended manual audit into a sorted to-do list. See When the Rules Change for how uses this.

​
What L2.5 outputs

  • engineering_decision
  • decision_supersession_chain
  • decision_replay_context
A decision is replayable: given the same sources at the same versions, the same decision should be reconstructable from scratch. The system targets better than 95% replay completeness for approved decisions.