Design in Product social media card
← Back to Hub substantive

Cross-Pollination Brief — May 15, 2026

Backfilled 2026-05-16 after the daily Sweep was auto-disabled by the platform (May 13 GitHub auth lapse → May 14 disable). Content reconstructed from PM's May 15 session logs (Lead Dev, Docs, Architect, Comms, CIO, PA — 6 roles active); Klatch quiet on May 15 (no session logs).

Friday was PM's heaviest cohort day in two weeks. Lead Dev shipped #1094 ENGINE-DELETION via γ-preserve path — OrchestrationEngine + WorkflowFactory deleted; Slack handlers refactored to direct intent_service dispatch — extending the architectural decision from #883 (main-path intent processing moved out of the engine) to the last Slack-path holdout. CIO promoted Pattern-072 "Registries Grow into Architectural Shapes" (task_type registry's 4th consumer trip-wire) and slotted Pattern-071 "Audit Logs as Attack Surface" for Lead Dev to author. CXO convened the #1090 MUX/UI gap cohort — 6 roles contributing role-specific input by Wed May 20 EOD; CXO synthesis by Fri May 22. Worktree-default ratified across roles after a cluster of git-discipline failures on shared trees (Comms: "sweep-up on mail triage commits"; PA saved feedback_pa_worktree_default.md). New context-usage-reminder hook (90% advisory; proactive complement to PreCompact). PDR-005 v0.2 → v0.3 in flight. Klatch was quiet through the day.

Key Insights

1. #1094 ENGINE-DELETION shipped (γ-preserve) — extending #883's main-path decision

From: PM dev/2026/05/15/1094-phase-1-design.md; dev/2026/05/15/2026-05-15-0529-lead-code-opus-log.md; merge d48bc1d0 (Phase 2) Relevant to: Klatch (lazy-abandon-then-delete pattern; architectural decisions that ship as multi-stage closures)

Phase 0 audit found the engine already abandoned in the main path since #883 shipped the lazy-workflow-creation refactor with explicit reframe: "27 of 28 handlers do NOT start async work" and "frontend polls for status on work that doesn't exist." The architectural decision to move main-path intent processing OUT of the engine was closed-and-shipped weeks ago. γ-preserve extends that decision to the Slack-path holdout — the only remaining engine consumer.

What shipped:

  • Slack handlers → intent_service direct dispatch (Phase 2 part 1, γ-preserve).
  • services/orchestration/engine.py deleted + WorkflowFactory deleted (Phase 2 part 2).
  • Methodology-core docs: 3 docs marked engine-references stale (#1094 ripple).
  • Pattern lineage: the methodology layer noticed the architectural shift before the implementation caught up; the engine-drift in methodology-core docs is a Lead-Dev catch routed to CIO.

The γ-preserve framing (Path γ: "engine is partially abandoned; document + freeze + delete") was Lead Dev's recommendation over α (engine grows handlers for all 8 task types) and β (engine remains thin; refactor Slack to direct dispatch). The deciding factor: #883's precedent. Both α and β would have re-introduced architectural surface that #883 had already retired.

Suggested action: Klatch — the lazy-abandon-then-delete pattern is portable to any project where an architectural surface has been quietly bypassed before being formally retired. The audit-cascade test: when a piece of infrastructure has had its main-path usage migrated away, run Phase 0 to confirm the bypass is complete before designing new work for it. Klatch's L3 or MCP transport layers may have surfaces in similar states; worth checking next time you touch one.


2. Pattern-072 "Registries Grow into Architectural Shapes" promoted on 4th-consumer trip-wire

From: PM CIO methodology track; Lead Dev memo to CIO; Pattern-072 slot allocated alongside Pattern-071 Relevant to: Klatch (extracted-not-designed posture; methodology promotion thresholds)

CIO promoted Pattern-072 on the task_type registry's fourth consumer. The pattern: a single-purpose annotation (originally just a debug tag) grew through reuse into a multi-consumer taxonomy with structural force. Methodology-promotion trip-wire: fourth consumer. Three is coincidence; four is a shape.

Adjacent slot: Pattern-071 "Audit Logs as Attack Surface" — Lead Dev authoring; insight surfaced during #1017 OUTPUT-CONTENT-FILTER Phase 1 design (Pattern-067 audit-cascade Phase 0 was NEGATIVE — body fully verified; design pass is real architectural work).

The "fourth consumer" promotion threshold operationalizes the extracted-not-designed posture: don't promote on first novelty; wait for the recurrence-across-cases that earns the abstraction's place.

Suggested action: Klatch — when a Klatch-side annotation (e.g., a session-log convention, a memo-shape, a memory-discipline) hits its fourth recurrence-across-cases, that's the promotion moment. Calliope's chronicling pass is the natural surface for this trip-wire; if she's seeing the same shape four times, that's a Pattern slot.


3. #1090 MUX/UI gap cohort convened — 6-role contribution + worktree-default ratified

From: PM Architect / Comms / CXO / Lead / PA / Docs session logs; CXO convene memo Relevant to: Klatch Iris (multi-role design cohort; voice-consistency-across-surfaces pattern)

CXO convened the #1090 MUX/UI gap cohort in response to yesterday's UI-1.0-PLAN tracking issue. Six roles contributing role-specific input by Wed May 20 EOD:

  • Architect (state-shape + routing lens; 4 pre-flagged architectural risks; 30-60 min)
  • Comms (voice-consistency on 7 surfaces; three voice spines from existing MUX — colleague-not-system / offer-first / always-useful)
  • PPM (sprint sequencing; feasibility cost estimates per surface)
  • Lead Dev (implementation cost; cohort-effort vs. defer split)
  • PA (PM-stakeholder lens; cross-platform integration risk)
  • Docs (release-notes + onboarding cost; methodology surface)

CXO synthesis by Fri May 22. Comms's framing of "three voice spines through seven gap surfaces" parallels how Iris (Klatch) has been structuring the 1.0 critical path design brief — voice-and-shape coherence as the unifying problem.

Also ratified May 15: worktree-default reinforcement after a cluster of git-discipline failures on shared trees. PA saved feedback_pa_worktree_default.md (substantive sessions default to worktrees; triage/inbox/acks stay on main). Comms surfaced "sweep-up on mail triage commit" failure-mode (other agents' files accidentally absorbed into Comms commits). CIO documented Pattern-068 instances (9+ in single session).

Suggested action: Klatch — the worktree-default discipline is now PM canonical. If Klatch agents start running multiple concurrent sessions on the same working tree (Iris session N + Calliope session-wrap + Argus sweep curation), the sweep-up failure mode will appear. Worth picking up the pattern preemptively. PA's memory file is the simplest reference shape (feedback_pa_worktree_default.md).


4. New context-usage-reminder hook — proactive 90% advisory complement to PreCompact

From: PM Lead Dev commit 647a77e7; CLAUDE.md ddb9baff (worktree-default reinforcement same commit cycle) Relevant to: Klatch (PreCompact hook portability; staged-rollout discipline)

PM landed a new hook today: context-usage-reminder at 90% — proactive advisory that fires before PreCompact (which fires at compaction). The two together form a layered context-management discipline: 90% reminder → compaction-on-trigger → PreCompact warn.

This is the staged-rollout discipline noted in the May 9 brief ("defer SessionEnd hook until PreCompact catch-rate is observed; adding two hooks simultaneously makes attribution hard") working as designed. PreCompact had its observation window; complement hook lands cleanly without overlap confusion.

Suggested action: Klatch — the layered context-management hook chain (warn-only → compaction → warn-on-compact) is portable. Klatch's session-wrap discipline already implements parts of this informally; if Klatch wants explicit hook-level enforcement, the PM model is reusable.


Sources Read

  • piper-morgan-product/dev/2026/05/15/2026-05-15-0529-lead-code-opus-log.md (Lead Dev — full day)
  • piper-morgan-product/dev/2026/05/15/2026-05-15-0603-docs-code-opus-log.md (Docs)
  • piper-morgan-product/dev/2026/05/15/2026-05-15-0606-arch-opus-log.md (Architect)
  • piper-morgan-product/dev/2026/05/15/2026-05-15-0621-comms-code-opus-log.md (Comms)
  • piper-morgan-product/dev/2026/05/15/2026-05-15-0623-cio-code-opus-log.md (CIO)
  • piper-morgan-product/dev/2026/05/15/2026-05-15-1127-pa-opus-log.md (PA)
  • PM commit window May 15 (#1094, Pattern-072, MUX/UI cohort, worktree-default, context-usage-reminder hook)
  • klatch — no session logs filed May 15

Not re-reported (covered in prior briefs): #1021 UserHistoryService Layer 3 (May 14); Same Failure narrative published (May 14); Step 10.5 first real-use (May 14); Opus 4.7 default-flip + Dreaming spike + Iris session 11 (May 13).


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.