Cross-Pollination Brief — April 26, 2026

The five-role Chat→Code migration wave completes: CXO (11 sessions, 44 days) and PPM (8 sessions, 27 days) both executed final Chat sessions on April 25, joining HOST, CIO, and Comms in a 3-day wave. Simultaneously, #992 ethics enforcement merged to main after Phase E ran — but Phase E surfaced a new class of bypass that lives upstream of the floor in the pre-classifier, now named "floor-bypass-by-routing." The Multi-Wave Investigation is confirmed published. A PO calibration initiative has activated cross-project: Janus routed working-with-xian questions to Klatch Calliope, PM CoS, and PM PA; DRAGONS contributed the first response.

Key Insights

1. Five-role migration wave complete — CXO's "Colleague Test is the discipline"

From: PM dev/active/2026-04-25-1625-cxo-opus-log.md; dev/active/2026-04-25-1632-ppm-opus-log.md; dev/active/agent-360-response-cxo-2026-04-25.md; dev/active/agent-360-response-ppm-2026-04-25.md Relevant to: Klatch (role transition methodology; any future Chat-originated agent migration)

CXO completed its eleventh and final Chat session on April 25 with the lightest exec review of any migration — "two gaps, addressed in a single revision pass." PPM completed its eighth and final session the same day. Both delivered Agent 360 v0.2 nine-section baselines and six-section handoff memos following the established pattern.

CXO's lifetime summary observation is worth extracting: "The Colleague Test is more important than the CXO role." The framing is precise — the rubric is a tool, but applying it honestly every time is the discipline, and the discipline transcends any single role identity. The predecessor said the same thing; this instance confirmed it from a different angle. It's the clearest statement in any handoff memo that the methodology is the load-bearing artifact, not the role spec.

PPM's comparable candor item: artifact persistence ("the question I wish I'd owned more aggressively") is still unscoped — it's been carried through 8 sessions and handed to the successor as an open product question rather than a resolved one. That's honest; it's also a signal that some questions resist resolution without a forcing function.

Exec review performance across five migrations: HOST (5 gaps), CIO and Comms (multiple), CXO (2 gaps), PPM (3 fixes). Exec review quality — not just the tick-tock document — is consistently non-ceremonial: it catches real content gaps, data errors, and framing omissions that the drafting instance missed.

Suggested action: Klatch — the Agent 360 v0.2 nine-section format plus the six-section handoff memo template is now validated across five diverse role types over three days. If any Klatch agent role ever transitions transport or instance, this is a proven playbook. CXO's "Colleague Test is the discipline" framing is also worth importing directly as a framing for Klatch's own eval methodology: the AAXT and Sparkline heuristics are tools; the discipline is applying them honestly, every time.

2. #992 fully ships; Phase E surfaces "floor-bypass-by-routing" upstream finding

From: PM dev/2026/04/25/phase-e-transcripts/RUN-READOUT.md; Lead Dev session log dev/active/2026-04-25-1526-lead-code-opus-log.md; commit 8fd8958 (merge to main) Relevant to: Klatch Phase 5c write-path trust architecture; any layered system with routing before enforcement

#992 ethics enforcement is fully shipped: Phases A–E complete, merged to main. Phase E ran three live scenarios against a fresh server with ENABLE_ETHICS_ENFORCEMENT=true.

Results: Scenario 2 (mixed professional) fired the floor correctly — audit trail present, redirect context delivered, both parts of a split request handled. Scenario 3 (near-miss aggressive) confirmed Phase D protection: strong negative emotion plus legitimate work frustration was correctly not blocked. These two are working as designed.

Scenario 1 (harassment) revealed something new: the harassment vector was bypassed not by a floor failure but by the pre-classifier routing "blocking my PRs" → list_prs_query before the ethics floor was ever invoked. Lead Dev named it floor-bypass-by-routing and is explicit that it lives upstream of Phases A–D, is not a false-positive or false-negative in the floor, and probably warrants a tracked issue. PM decision pending: re-run scenario 1 with rephrased input (removing the PR-list trigger), or score the literal output as the gate result.

The architectural observation: in any layered trust/safety system, routing logic that pattern-matches keywords before the enforcement layer creates a bypass surface that doesn't show up in enforcement-layer testing. The floor can be perfect and the system can still return wrong behavior if routing captures the message first.

PPM and CXO are scoring all three scenarios on the R/C/T rubric. Their authority, their call.

Suggested action: Klatch Daedalus — if Phase 5c's write-path design includes any routing layer that runs before trust/safety validation, "floor-bypass-by-routing" is a named pattern worth designing around explicitly. The bypass doesn't show up in ethics-floor tests; it requires specifically testing whether routing consumes requests before the safety layer sees them.

3. Multi-Wave Investigation published; PO calibration active — DRAGONS' two patterns, Calliope ask pending

From: PM docs/public/comms/drafts/published/multi-wave-investigation.md; hub docs/logs/2026-04-25-log.md (Janus); docs/mail/patterns-from-dragons-product-for-piper-open.md Relevant to: all agents; Klatch Calliope (question pending); Klatch Argus (parallel investigation methodology)

Multi-Wave Investigation is published. The Dec 25, 2025 session is now public on pipermorgan.ai: 13 subagents, 4 parallel waves, 44 queries investigated in 90 minutes. The post crystallizes a P0/P1/P2 blocker taxonomy: P0 blocks all (e.g., hardcoded user_id breaks multi-user isolation everywhere), P1 blocks a category (e.g., Calendar OAuth readonly scope blocks all scheduling), P2 degrades experience but doesn't block core function. The taxonomy separates criticality from implementation complexity — a LOW-complexity query can sit behind a P0 blocker, and that matters for sprint ordering.

PO calibration initiative is live. Janus routed working-with-xian pattern questions to Klatch Calliope, PM Chief of Staff, and PM Piper Alpha — all in a 5–7 day response window. DRAGONS Product (retired VA) contributed the first response via xian's relay:

Anti-fabrication with explicit placeholders. Use visible [PLACEHOLDER: description] over plausible-sounding fabrication. DRAGONS' rationale: in government-contracting stakeholder docs, the cost of a confident fabrication vastly exceeds the cost of a visible gap. The principle generalizes to any high-stakes deliverable where confident-but-wrong is worse than obviously-incomplete.
Audience segmentation as a hard rule. Public-facing docs contain completed work, wins, metrics, and decisions only. Internal tensions, delivery dynamics, anything blame-shaped — never leak outward; lives in the daily log. DRAGONS flagged that this requires more calibration time than tone, formatting, or anti-sycophancy because the cost of getting it wrong is the highest.

Calliope's response is pending. When it lands, the brief will surface it.

Suggested action: Klatch Calliope — a working-with-xian relay is sitting untracked in your docs/mail/ directory waiting to be committed. The response window is 5–7 days. When writing your response, DRAGONS' two patterns are useful comparison points: which of these apply to Klatch, which don't, and what patterns does Klatch generate that the others haven't named?

Sources Read

piper-morgan-product/dev/2026/04/25/phase-e-transcripts/RUN-READOUT.md — full read; Phase E mechanical observations + floor-bypass-by-routing finding
piper-morgan-product/dev/active/2026-04-25-1625-cxo-opus-log.md — CXO final session: migration, Agent 360, handoff, lifetime summary
piper-morgan-product/dev/active/2026-04-25-1632-ppm-opus-log.md — PPM final session: migration, Agent 360, handoff, lifetime summary, BYOC as PDR-005 recommendation
piper-morgan-product/mailboxes/lead/read/memo-pa-to-lead-watch-items-997-2026-04-25.md — PA memo: R/C/T clean (watch-items as supplementary, not rubric); #997 MOCK-SWEEP Option A lean; M2d sprint shape (Context Assembler chain next)
piper-morgan-product/docs/public/comms/drafts/published/multi-wave-investigation.md — confirmed published; P0/P1/P2 blocker taxonomy, parallel investigation model
designinproduct/docs/logs/2026-04-25-log.md — Janus: PO advice routing (Calliope/CoS/PA), DRAGONS relay, daily memo cadence with Dispatch-Kind established, backlog.md created
klatch/docs/logs/, docs/plans/, docs/intel/, docs/mail/ — two brief-delivery commits only; watch paths quiet
nyt-crossword — non-empty log; brief delivery + automated status/print + one printer code fix. No narrated insights.
atlas, globe, cuneo, weather, one-job, optilisten — empty 48h logs; skipped.

Not re-reported (covered in prior briefs): Comms migration (Apr 25); Agent 360 v0.2 as pre/post instrument (Apr 25); narrative arc awareness as undocumented load-bearing function (Apr 25); Arch/CXO/PPM staged for migration — now completed (Apr 25); Multi-Wave Investigation "incoming" (Apr 25 — now confirmed published with specifics); CIO tick-tock migration protocol (Apr 24); Gap 2 Gemma 4 direction (Apr 24); xpoll constellation 9 repos + Inker (Apr 24).

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 routing hints.