Skip to main content
Everything the substrate learns — cortex lessons, forge remember facts, verified reuse artifacts — lands as content-addressed claims in a git-native ledger (.forge/ledger/) built to merge without conflicts. There is no server and no sync service; it is just files in git.

​
Team memory in three commands

1

Initialize once

Among other things this emits the .gitattributes union-merge rule the ledger needs.
2

Work normally

Cortex lessons and forge remember facts shadow claims into the ledger as you go — nothing extra to run.
3

Fold in a teammate's ledger

Any order — the merge is conflict-free.

​
Why it can’t conflict

A claim’s bytes are a pure function of (kind, body, scope), so every replica computes the same identity for the same knowledge. The merge is a join-semilattice — property-tested to be commutative, associative, and idempotent — so two teammates’ ledgers converge to the same state no matter who syncs first.
Identical knowledge minted independently converges to one claim with every author preserved in its provenance.

​
Trust and provenance

Confidence is moved only by independent oracles — tests, CI, a human accept/revert — so importing a teammate’s ledger doesn’t blindly trust their notes; it imports their evidence.

​
Reuse across the team

Once a teammate’s verified code is in the merged ledger, you can reuse it with its proof:
A hit points at working, test-confirmed code and the forge ledger blame that proves it — reuse it instead of regenerating.
Dormant claims are kept for audit, never deleted; unreviewed knowledge decays toward uncertainty, not toward deletion. The ledger is an evidence trail, not a cache you can silently lose.