This hub gathers how change flows through aswritten's repo-native operations — across handoffs, research promotion, review rounds, and branch policy. On process: [handoffs are never documents] (Claim_HandoffsNeverDocuments) (Jun 2026), living in task notes and the ledger, with the 811-line launch-plan blob as the named failure mode. On research-to-canon, [promotion is rewriting, not merging] (Claim_PromotionByRewriting) (Apr 2026): the research branch is raw material, not a merge target. Review mechanics recur twice — [the corrections-vs-merge semantics] (Claim_ReviewMemorySemantics) and [the same flow as narrative] (Narrative_ReviewMemoryFlow) — both holding that corrections take at least two rounds. Branch policy is [git merge into dev, never rebase] (Obs_GitMergePolicy), with squashing flagged as problematic. The [Notion retirement narrative] (Narrative_NotionRetirement) (Jun 2026) makes explicit a long-abandoned tool's formal end, the first Organization-domain memory, with [Scarlet's Mar 2026 merge-constraint quote] (Observation_ScarletDevMergeQuote) marking an in-flight pause on new pushes.
Witness phrases
verbatim Handoffs are never documents.
verbatim Promotion is rewriting memories into main's voice and structure at the moment of decision — not cherry-picking commits or merging the branch.
verbatim A review with corrections always takes at least two rounds.
verbatim Use git merge main to pull changes into dev; never rebase.
verbatim CRM migration (task-268) is the last Notion dependency.
verbatim I'm trying to get all of this stuff merged from dev domain, but we're not doing any new pushes.
verbatim Two opposed change-flow models coexist: research changes are rewritten into canon (never merged), while dev changes are pulled in via git merge — promotion vs merge are deliberately different mechanisms.
verbatim Notion retirement is asserted as complete yet qualified: CRM migration (task-268) is still an open dependency, so 'fully retired' is aspirational, not finished.
verbatim Scarlet's Mar 2026 quote signals a freeze — merging from dev is in scope but no new pushes — a temporal constraint that sits against the otherwise forward-moving merge/promotion process.
verbatim Review semantics appear in duplicate (a Claim and a Narrative carrying the same corrections-vs-merge rule), one tagged aboutNode rather than broader — possible redundancy or a node-level vs narrative-level distinction.
In this area
ClaimClaim_PromotionByRewriting Promotion is rewriting memories into main's voice and structure at the moment of decision — not cherry-picking commits or merging the branch. The research branch becomes a source of raw material, not a merge target. This matches the pattern
ClaimClaim_ReviewMemorySemantics Correction 7: Save memory = corrections exist, ends round, PR updates. Merge = no corrections needed, closes PR. A review with corrections always takes at least two rounds.
NarrativeNarrative_ReviewMemoryFlow Saving a review memory ends a round and triggers extraction; merge signals no corrections needed and closes the PR — corrections always require at least two rounds.
4 further nodes are not published in the public projection. Each node links from the entity index.