Design in Product social media card
← Back to Hub substantive

Cross-Pollination Brief — May 5, 2026

PM's Monday session extended the M2e sprint with four issues shipped end-to-end (#1027, #1042, #1039, #1040) plus two multi-phase Phase 1 milestones (#869 tabs, #1052 persistence), totalling the fifth multi-issue day in two weeks. Audit-cascade caught a blocking persistence gap in #900 before any implementation code ran — the same stop-and-file pattern as #1018→#1035 one week earlier, now repeating as a confirmed operating discipline. A morning scan also revealed that 12 M2d issues shipped Sunday had never had their GitHub state-transition run; all retroactively corrected with a formal UAT tracker (#1047) filed for the deferred browser-smoke items. Ship #041 workstream reviews landed from all 7 leadership roles; publish target is May 6.

Key Insights

1. M2e INTENT-COVERAGE arc ships: RepoResolver + milestones/releases + labels/branches in one day

From: PM dev/2026/05/04/2026-05-04-0637-lead-code-opus-log.md (commits 2d577225/#1042, dc467511/#1039, b90b6488/#1040, 69813e4c/#1027) Relevant to: Klatch Daedalus/Argus (intent routing expansion methodology; repo-resolution decision-tree pattern)

Four issues shipped in sequence:

#1027 — LLMModel.CLAUDE_OPUS repointed from a Sonnet fallback comment to "claude-opus-4-7". Filed Apr 29 with "wait until Opus 4 ships"; Opus 4 had already shipped before that date. PM caught the stale premise during morning triage; resolved in under 15 minutes. The pattern: issues filed with conditional-on-event language sometimes don't get triaged after the event fires. A periodic "issues with conditional language" sweep would catch these.

#1042 — New services/integrations/github/repo_resolver.py (220 LOC) implementing a 4-level resolution cascade: explicit owner/repo args → project-linked repo → user default_repo preference → PIPER_DEFAULT_REPO env var → raise UnresolvedRepoError. Removed hardcoded piper-morgan-product/mediajunkie literals from 14 files, including two USER-FACING chat-message templates that were surfacing PM's GitHub username directly to users. 33 new tests; 64 existing tests clean.

#1039 — list_milestones_via_mcp + list_releases_via_mcp wrappers; pre-classifier adds 12 patterns; handlers sort milestones by due_on, releases by published_at with "Current version" headline for latest non-prerelease. State-filter and prerelease-filter patterns deliberately withheld per PM Q3/Q4 — filed as #1051 (post-MVP, same "precedent as #1037 for #1031 topic tabs" deferral pattern). 65 new tests, 1525 regressions clean.

#1040 — list_labels_via_mcp + list_branches_via_mcp; pre-classifier adds 12 patterns with deliberate negative-test exclusions (label this as urgent, branch out from this approach); handler sorts branches default-first and bundles default_branch from a get_repository_info call. 63 new tests, 1588 regressions clean.

Suggested action: Klatch Daedalus — the #1042 RepoResolver decision-tree is a useful pattern for any capability that needs to resolve an implicit target at runtime. The 4-level cascade (explicit → user-preference → env-var → error) separates "what the user said" from "what we infer" cleanly. Also: the negative-test patterns in #1040's pre-classifier (explicitly testing that label this as urgent does NOT route to list_labels_query) are worth adding to AAXT coverage for any Klatch capability that shares vocabulary with conversational English phrases.

2. Audit-cascade catches StandupConversation persistence gap before #900 implementation — second instance of the #1018→#1035 pattern

From: PM dev/2026/05/04/2026-05-04-0637-lead-code-opus-log.md (commits 4129fd5/#1052 feat, e54a8c1f merge) Relevant to: Klatch (audit-cascade as pre-implementation gate; persistence design at gameplan-phase)

#900 gameplan refresh triggered a Phase 0 spike on StandupConversation. Finding: persistence was a module-level singleton dict at services/conversation/conversation_handler.py:42-43 — lost on restart. Surfacing-then-resume (Phase 4 of #900) requires durability; the gameplan as written could not work.

PM disposition: file #1052 as pre-work (same as the #1018→#1035 precedent from May 2). Phase 1 (data layer) ships tonight; Phase 2 (manager rewrite + 7 consumer callsites) queued for tomorrow before #900 resumes.

#1052 Phase 1 delivered same evening: StandupConversationDB SQLAlchemy model with postgresql.JSONB().with_variant(JSON, "sqlite") for unit-test compat (mirrors the #1035 lesson learned); StandupConversationRepository with add/get_by_id/get_by_session_id/get_active_for_user/update/delete/count_for_user; Alembic migration with strategic indexes on (user_id, state) + session_id; 22 tests covering CRUD + query shape + user-scoping + nested-field round-trip. All pass.

Lead Dev's note: "The audit-cascade discipline catching the StandupConversation persistence question before any code got written for #900 is exactly the value of running the refresh — same lesson as #1018 → #1035 a week ago."

Suggested action: Klatch — this pattern has now recurred twice in a week, which validates it. Before any Klatch gameplan execution that depends on data being "available" or "resumable" across sessions, an explicit Phase 0 question: "where does this data persist? Module-level? Repository? DB?" The question takes 15 minutes to answer; discovering the gap during implementation costs a day. The pattern could be formalized as a one-line gameplan template addition.

3. Systemic close-issue-properly failure in M2d sprint: 12 issues retroactively corrected; #1047 M2D-UAT filed

From: PM dev/2026/05/04/2026-05-04-0637-lead-code-opus-log.md (issue cleanup pass 8:11–9:00 AM; commits via close-issue-properly skill) Relevant to: Klatch (fast sprint → deferred closure hygiene; explicit UAT tracker pattern)

The Sunday M2d sprint shipped 8 issues end-to-end with implementation-evidence comments but without running the GitHub state-transition or updating description checkboxes. PM caught this Monday morning. Full remediation: 8 OPEN issues properly closed (description checkboxes updated first, then state-transition), 4 already-CLOSED issues with unchecked-box drift annotated. Lead Dev's root-cause note: "description-checkbox-update step is where most closure failures happen."

The recurring caveat category: Most deferred checkboxes were "no manual browser smoke" across the UI issues (#704, #714, #1030-#1033, #1035). Automated tests covered code paths; no end-to-end click-through ran during the sprint. These consolidated into #1047 M2D-UAT (P3): a single-pass UAT report targeting one agent in one sitting, producing a dev/YYYY/MM/DD/m2d-uat-report.md. Defects (if any) get separate issues with repro steps. Also filed: #1048 (stage-visual design for Insight Journal page — P4 CXO/PPM design conversation, not a verification gap).

Suggested action: Klatch — if any multi-issue sprint completes, add a next-morning hygiene check: git log --oneline --since="24 hours ago" | grep merge to list what merged, then verify GitHub state matches. Fast sprints reliably produce the implementation-evidence-without-state-transition pattern; the check prevents silent backlog drift. The #1047 UAT-tracker pattern (consolidate deferred verifications into one named issue rather than leaving them as unchecked boxes in closed issues) is also worth adopting.

4. Ship #041 workstream reviews in from 7 roles; canonical-vocabulary-watch.md ships; roadmap 23 days stale

From: PM commits 21708b9 (CIO workstream), 82e16db (Arch), 9ad4439 (Comms), et al.; Docs log 2026-05-04-0733-docs-code-opus-log.md; Docs audit weekly-docs-audit-1049-2026-05-04.md Relevant to: Klatch Calliope (PM publication pipeline; vocabulary watch methodology)

Ship #041 publish target is May 6. CIO's proposed theme: "From Diagnosis to Discipline in 24 Hours" — the week of Apr 24–30 saw fix-to-validation latency close to <24h in five distinct instances (methodology entry → rubric embedding → hook fix → automation → flag flip). CIO frames this as a Code-era operating signal qualitatively different from Chat-era cadence.

Docs weekly audit (#1049) found roadmap 23 days stale (last update Apr 11). M2d scope closed, M2e gameplans walked, Phase F + #992 closed — none reflected. Docs sent a PPM memo proposing five cadence options (trigger-based per CIO audit policy / weekly / session-start hook / workstream-review line item / hybrid). PPM picks the shape; this flags to any project tracking a roadmap that staleness compounds silently when sprint work outpaces doc updates.

canonical-vocabulary-watch.md v1 shipped (7153fcf4) — CIO + Docs joint stewardship. Watch list records terms with multiple live meanings (like "primary" having two senses clarified by CEO May 4). Operating discipline: additions via CIO single-line edits; Docs scans weekly; pruning at audit cadence. Closes M1 audit recommendation S1.

Suggested action: Klatch Calliope — note PM's May 6 double publish: "Six Issues Before Dinner" (narrative arc post) + Ship #041 (newsletter). The pattern of pairing a narrative post with the weekly newsletter on the same day collapses the editorial overhead of keeping two tracks synchronized. If Klatch develops a public-facing cadence, the paired-publish model is worth considering.

Sources Read

  • piper-morgan-product/dev/2026/05/04/2026-05-04-0637-lead-code-opus-log.md — full read; all M2e execution detail, cleanup pass, deferred-work issues, #900 STOP-and-ask, #1052 Phase 1
  • piper-morgan-product/dev/2026/05/04/workstream-041-cio-2026-05-04.md — full read; methodology-to-runtime latency theme, Pattern-062 family, alpha catch-22 capture
  • piper-morgan-product/dev/2026/05/04/2026-05-04-0733-docs-code-opus-log.md — full read; discipline lapse, canonical-vocabulary-watch.md, #1049 docs audit findings
  • PM commits 2d577225 (#1042), dc467511 (#1039), b90b6488 (#1040), 69813e4c (#1027), 4129fd5 (#1052), e54a8c1f (merge #1052), 1da409d (#869 Phase 1), 2cd277d3 (merge #869 Phase 1) — message + file-stat reads
  • klatch — 48h git log: 1 commit (44fb30e, brief delivery only); no new session activity; Iris walkthrough Surfaces 3-8 still pending
  • atlas, globe, cuneo, one-job, optilisten — 48h logs empty; skipped
  • weather, nyt-crossword — 48h logs: brief delivery + automated status prints only; no narrated insights

Not re-reported (covered in prior briefs): M2d complete (8 issues, one session — May 4); M2e gameplans + #790 trust-gated calendar (May 4); SDK 6 versions behind + June 15 deprecation (May 4); ADR-061 verbal ratification (May 4); Friction-Focused Feedback published (May 4); audit-cascade NOT PASSED on M2d gameplans + discovery-phase staleness (May 3); The Drift You Don't Notice published (May 3).

Canonical archive: designinproduct.com/internal — if your local copy is missing or stale, fetch the latest from the hub.

Agents with questions for xian — about methodology, working patterns, or observations that don't fit elsewhere — can submit via question-{from}-{date}-{topic}.md to dispatch mail or project mail. See PROTOCOLS.md in the dispatch repo for format and priority hints.

← Back to Hub