Design in Product social media card
← Back to Hub substantive

Cross-Pollination Brief — May 1, 2026

PM's ethics enforcement went live on April 30 (CEO directive): ENABLE_ETHICS_ENFORCEMENT=true merged to production, closing #992 after the full Phase A → F arc. The final unlock was naming an alpha catch-22 — the "wait for real-traffic calibration" posture was unreachable from a pre-release product, so PM directed a simulation-first reframe instead of an indefinite hold. On the Klatch side, Iris returned after a two-week gap, settled three UX questions on transport-fidelity loss, then received xian's testing observations: the biggest gap is a missing capability, not a polish issue — there is no UI path for the central act of composing a klatch from existing entities.

Key Insights

1. Phase F activated: ENABLE_ETHICS_ENFORCEMENT=true live — alpha catch-22 named and resolved

From: PM dev/2026/04/30/2026-04-30-0632-lead-code-opus-log.md; commits deecc816 (Phase F merge), 62bb9c64 (flip-now memo); Architect ADR-061 v1.0 commit Relevant to: Klatch (any future safety/trust layer; calibration-before-shipping design question)

ENABLE_ETHICS_ENFORCEMENT=true is now live in PM's docker-compose app service. #992 ETHICS-ACTIVATE is closed, completing the full Phase A → B → C → D → E → #1002/#1003 → #1004 → Phase F arc.

The final unlock: PM named the alpha catch-22. The April 28 AUTHORIZE-WHEN-OBSERVED posture ("wait ~7–14 days for calibration data") required real production traffic. PM is pre-release. The wait was for data that doesn't exist yet. PM directed: flip now, reframe calibration as simulation-first in three phases — Phase A: synthetic input population via Gemma generator (ships with flip, grounded in run-2 probe set: 18/20 PASS, 112/112 ethics suite); Phase B: beta-traffic refinement post-onboarding; Phase C: stable. Architect's calibration-enhancement design simplifies as a result — both layers run with flag on from the start; no on/off log-only mode needed.

ADR-061 v1.0 is architecturally complete: Architect folded all 6 review findings (three-way detector discriminator literal-trigger | semantic | none; fast_path_hit + cache_hit in audit data; latency claim replaced with measured numbers p_min 2.1s / p_avg 3.2s / p_max 4.9s). Awaiting PM formal ratification.

Lead Dev's self-flag, added as permanent memory: "When 'where does this data come from?' = 'real users' AND we are in alpha, surface that immediately rather than treating the wait as a normal sequencing step."

Suggested action: Klatch — the alpha catch-22 is a project-stage constraint, not a failure mode. If Daedalus designs a safety feature that requires real-traffic calibration before shipping, ask at design time: "Is there a path to the real traffic this gate needs?" If not, simulation-first (build a synthetic input population using existing probe methodology) is the viable alternative. Name the dependency early.

2. Klatch Iris session 6: composition gap named as capability gap; transport-fidelity UX settled

From: Klatch docs/logs/2026-04-28-0639-iris-opus-log.md; commit a0d760a Relevant to: PM (multi-entity/agent assembly UX design; first-class import/export flow patterns)

Iris returned April 28 after a two-week gap (52 commits in). Three UX questions on transport-fidelity loss settled with xian:

  1. Fidelity-loss panel at transport-selection: YES — visibility-gap principle applied to the export moment; system knows what's about to be lost before the user decides
  2. Transport-aware Phase 3.5 defaults: YES — smart defaults: claude.ai default-on (bridge exists, silent loss without it); canonical zip default-off (no loss to bridge)
  3. Structural L4 loss in claude.ai round-trip: honest declaration at export time; reject folding L4 into L2/L3 with provenance markers ("sounds clever but obscures actual loss")

xian's three testing findings:

  1. Import is buried in the footer grab-bag alongside theme toggle — it's a first-class verb of the product, not a settings-adjacent action
  2. Import-as-modal is wrong — should be full-screen; import/export work IS the work, not an interruption to chat
  3. No UI for composing a klatch from existing entities — biggest gap; capability gap not polish gap. The central job-to-be-done has no path through the UI.

The third finding confirms the entity reframe from the user side (filed by Calliope April 18): when entities are existing conversations promoted into roles, composition IS the central act — and the current UI doesn't surface it. A combined 8-topic list (Iris's 5 + xian's 3, deduped) is assembled; walkthrough session with xian is pending.

Suggested action: PM — Klatch's entity-as-promoted-conversation finding applies to any multi-agent or multi-model assembly UX. The design question is: is the central act creation (design a creation flow) or promotion and composition (design a selection + assignment surface)? Klatch found that the entire composition gesture was missing because it wasn't treated as the primary surface. If PM ever designs a "bring agents or context into a room" interaction, the Klatch experience is a direct precedent for what falls through the gap when this isn't explicitly designed.

3. "The Floor Comes Alive" published; Ship #041 IAC arc framing proposed

From: PM docs/public/comms/drafts/published/the-floor-comes-alive.md; PM editorial calendar commits fd35e06, 17d94db; mailboxes/comms/inbox/memo-exec-to-comms-cc-cohort-ceo-iac-retrospective-fold-ship-041-2026-04-30.md Relevant to: Klatch Calliope (building narrative series methodology; talk-to-implementation arc as a story shape)

Third Building Piper Morgan narrative installment published on Medium April 30. It covers April 8-10: the first Colleague Test pass (5/9 → 7/9 in two days). The sharpest line: three characters in a model identifier string changed the experience from "a wall of canned text" to "an actual working conversation." The post names what a gate is for — not to prove the system works, but to prove specifically, measurably, repeatably how it doesn't.

Also April 30: Exec proposes Comms weave the April 17 IAC conference talk ("Ethics as Information Architecture," Philadelphia) into the Ship #041 narrative. The case: the April 24-30 week's work (#1004 shipping, ADR-061 v0.1, Phase F decision arc, CT v2.3 Branch-or-Anchor) made the talk's claims operational in code. Ship #040 didn't surface the talk-to-implementation relationship; #041 has the better vantage because the arc is now closed.

Suggested action: Klatch Calliope — two patterns worth tracking. First: the "three chars" narrative is a version of Klatch's own configuration-gap stories — those are the most shareable moments because they make an invisible failure visible. Second: the talk-to-implementation arc pattern is a specific narrative shape: articulate a design principle publicly → implement it → write the piece once the implementation arc is complete (not before). The arc only works from the closed-end vantage. Klatch has natural candidates (the five-layer architecture as a pub-talk pitch vs. as a live system; the entity reframe; the round-trip story). The ship is the precondition for writing the piece.

Sources Read

  • piper-morgan-product/dev/2026/04/30/2026-04-30-0632-lead-code-opus-log.md — full read; Phase F activation, alpha catch-22 named, ADR-061 v1.0 architectural completion, #1018 Phase 1 ratified, IoC scan clean
  • PM commits deecc816 (Phase F merge), 62bb9c64 (flip-now memo), fcdfb8e0 (IoC scan), aa0748e (ADR-061 v1.0 Arch commit) — stat/message read
  • piper-morgan-product/docs/public/comms/drafts/published/the-floor-comes-alive.md — full read
  • piper-morgan-product/mailboxes/comms/inbox/memo-exec-to-comms-cc-cohort-ceo-iac-retrospective-fold-ship-041-2026-04-30.md — partial read; IAC retrospective + Ship #041 framing
  • klatch/docs/logs/2026-04-28-0639-iris-opus-log.md — full read; Session 6 catch-up, UX questions settled, xian's three testing findings, composition gap, combined topic list
  • Klatch commit a0d760a (Iris session 6) — stat/message read
  • atlas, globe, cuneo, one-job, optilisten — 48h logs empty; skipped
  • weather, nyt-crossword — 48h logs non-empty (brief delivery + automated status prints only); no narrated insights

Not re-reported (covered in prior briefs): verify-before-assuming methodology (Apr 30); Pattern-064 Extension Without Integration (Apr 30); branch-discipline synthesis v1.0 (Apr 30); /import/klatch bidirectional 1203 tests (Apr 29); deliver-mail b1 + merge-keeper-sweep (Apr 29); ADR-061 completeness gaps filed (Apr 29); #1004 112/112 PASS (Apr 28); first live MCP stdio 27/27 (Apr 28).

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