Piper Morgan Roadmap v15.0

Date: April 11, 2026 Author: PA, with leadership review (PPM, CXO, CIO, Architect) Status: Active. M1 closed April 11. M2 begins now. Supersedes: v14.3 (March 10, 2026) — archived at docs/internal/planning/historical/roadmap-v14.3-2026-03-10.md

Executive Summary

M1 (MVP Foundation) closed April 11, 2026 at 7/9 Gate 1 PASS after four UAT rounds. The closure follows a major strategic pivot: M2-M5 has been restructured around the differentiator stack (context methodology + conscious floor + artifact persistence + trust-graduated experience) per Vision V2.3. The previous sprint themes (Activation, Skills, Documents, Polish) are retired in favor of themes that map directly to what makes Piper Piper.

Key changes from v14.3:

• M1 closed — engineering complete, Gate 1 7/9 PASS, conversation continuity (#922) deferred to M2
• M2-M5 restructured around differentiator stack themes
• M6 retired — contents redistributed to M5 or Fast Follow
• 12 issues closed as superseded by intervening work (per backlog deep review Apr 7)
• 2 issues replaced with sharper scope: #312 (Design System) → Distribution Visual Identity (M5), #241 (Ethics Tuning) → Floor-First Ethics Verification (M2)
• 10 new issues filed (#950-959) covering differentiator stack work
• DIST moved earlier — distribution is part of M5, not after it. MCPB packaging is the primary distribution path.
• Distribution philosophy: Bring Your Own Chat — Piper as MCP server consumed by any MCP-compatible client

Current Position

Inchworm Position: 4.4.3 (M1 closed, M2 starting)

1. ✅ The Great Refactor (GREAT)
2. ✅ CORE functionality
3. ✅ ALPHA testing (v0.8.0 → v0.8.5.1)
4. 🎯 Complete build of MVP
   4.1. ✅ B1 - Beta Enablers (v0.8.3.x) - COMPLETE Jan 11
   4.2. ✅ A20 - Alpha Testing round 2 (v0.8.4.x) - COMPLETE Jan 18
   4.3. ✅ MUX: Modeled User Experience - COMPLETE Jan 27
   4.4. 🎯 MVP Sprints (M0 → M5) ← CURRENT
        ✅ M0: Conversational Glue — COMPLETE (v0.8.6, Mar 4)
        ✅ M1: Foundation — COMPLETE (gate closed Apr 11, 7/9 Gate 1 PASS)
        🎯 M2: Conscious Floor + Action Handlers ← STARTING
        ○ M3: Artifact Persistence
        ○ M4: Trust + Learning
        ○ M5: Distribution + Polish
5. Beta testing on 0.9
6. Launch 1.0

The Differentiator Stack (Vision V2.3)

The MVP sprints are organized around four differentiators that, together, make Piper a colleague rather than a chatbot wrapper:

1. Context Methodology — Five-layer model operationalized as practiced discipline
2. Conscious Floor — LLM responses embodying grammar, Five Pillars, anti-flattening
3. Artifact Persistence — Conversation outputs that outlive the conversation, with composting lifecycle
4. Trust-Graduated Experience — Earned proactivity through demonstrated value

Indoor plumbing (commodity, not differentiating): GitHub/Slack/Calendar/Notion via MCP plugins, file storage via SQLite, auth via standard OAuth, LLM provider management via adapter pattern.

MVP Sprint Status (April 11, 2026)

M0 — Conversational Glue ✅ COMPLETE

Status: COMPLETE — v0.8.6 shipped March 4, 2026 Theme: Natural conversation flow Final Totals: 27 issues (5 planned + 22 discovered via Assembly Assumption)

M1 — MVP Foundation ✅ COMPLETE

Status: COMPLETE — Gate #926 closed April 11, 2026 Theme: Floor-first routing + security + testing foundation Final Result: Gate 1 7/9 PASS (Q2 marginal, Q8 #922 conversation continuity deferred to M2) Test Suite: 6,309 passing, 0 failures

Key shipped:

• ADR-060 Floor-First Routing
• ADR-059 Workflow Dispatcher (onboarding removed per Gall’s Law)
• PDR-004 Experience Philosophy
• #940 LLM provider config simplification (single-provider, no hardcoding, error classification)
• #942 Missing orchestration tables
• #939 Avatar fixes
• #943 GitHub pre-flight checks
• Conversation continuity fix (storing agent responses in turn history)
• Floor fabrication guardrail (“NEVER list or invent user data unless explicitly in context”)

Lessons learned (for M2 planning):

• Pattern-045 (Green Tests, Red User) is systemic — three separate areas during gate testing
• Silent failures delayed diagnosis — graceful fallbacks must not be too graceful
• Server restart reliability (#949) cost real debugging time
• Quality testing methodology (Colleague Test) is the most impactful CXO contribution

M2 — Conscious Floor + Action Handlers 🎯 CURRENT

Theme: Make the conscious floor reliable and the action gate clean. The differentiator stack’s first two pillars become operational. Status: Starting April 11

Sprint focus: Floor reliability, context assembly across all floor-routed categories, action gate routing for the few side-effect handlers, E2E + AAXT testing infrastructure.

Issues:

Pre-sprint actions (per leadership review):

• File Floor-First Ethics Verification issue (replaces closed #241) — verify floor pipeline ethics matches pre-ADR-060 coverage
• Scope context assembler (#951) with explicit data sources — PA + PPM + Architect collaboration before implementation starts
• Address #949 (server restart reliability) early to prevent gate cycle contamination
• CXO reviews floor system prompt (#950) at sprint start
• Run E2E/AAXT tests (#927-930) from sprint start, not as final verification

Sprint expansion expectation: M0 expanded 5→27 issues (5.4x), M1 expanded ~2x. The context assembler is the kind of work that reveals gaps. Plan for 6-8 weeks rather than 4.

M3 — Artifact Persistence

Theme: Conversation outputs that outlive the conversation. The bridge between “good conversation” and “useful PM tool.” Status: Pending M2 closure

Issues:

Pre-sprint actions:

• CXO reviews artifact lifecycle data model (#952) — composting is experience design, not just storage schema
• Scope cross-session memory persistence (#953) tightly: what persists, what expires, how it enters the context window
• Composting data model designed in M3; composting engine deferred to M4

M4 — Trust + Learning

Theme: Earned proactivity and cumulative understanding. The differentiator stack’s third and fourth pillars. Status: Pending M3 closure

Issues:

Pre-sprint actions:

• Architect review of trust graduation implementation approach (lightweight via context vs. dedicated scoring service)
• PM emphasized: trust graduation MVP must be credible first step toward full model, not throwaway hack
• Composting engine implementation (deferred from M3)

M5 — Distribution + Polish

Theme: Get Piper into users’ hands via Bring Your Own Chat. Distribution earlier rather than after polish. Status: Pending M4 closure

Issues:

Pre-sprint actions:

• File Distribution Visual Identity issue (replaces closed #312) — icon, name, description, MCP Apps canvas design
• Verify Managed Agents docs and integration path (Claude Managed Agents launched April 8 as additional distribution target)

Closures and Revisions (Per Backlog Deep Review, April 7)

Closed (12 issues, all superseded by intervening work)

• #167, #191, #273, #276 — Old testing issues, superseded by E2E/AAXT track (#927-930)
• #146, #147, #148 — FLY-VERIFY trilogy, superseded by methodology (Completion Discipline Triad, audit-cascade, mailbox system)
• #309, #315 — Skills framework (CONV-MCP-PROTO, CORE-SKILLS-LIBRARY) — framework not adopted
• #313 — File Browser UI — #355 is the right-sized version
• #241 — Ethics Tuning — replaced by Floor-First Ethics Verification (M2)
• #312 — Design System — replaced by Distribution Visual Identity (M5)
• #118 — Multi-Agent Coordinator — superseded by document-based methodology
• (some closures may have been postponed rather than closed per PM judgment)

Revised (3 issues, scope updated)

• #100 CONV-FEAT-PROJ → M2 context shape (project portfolio data assembled for floor)
• #101 CONV-FEAT-TIME → M2 context shape (temporal context injection for floor)
• #310 CONV-UX-QUICK → Fast Follow (post-MVP UX polish, P0 demoted)

Deferred to Horizon 2

• #103 CONV-FEAT-PRIOR (Priority Engine) — conversational reasoning via floor is MVP-sufficient

Distribution Strategy: Bring Your Own Chat

Build sequence (Gall’s Law):

1. MCP server — standalone, runs via stdio transport, exposes Piper tools (Python recommended per Architect for codebase reuse)
2. MCPB packaging — bundle for Claude Desktop one-click install
3. Claude Project template — persona/instructions for the hybrid approach (resolves system prompt injection gap)
4. MCP Apps — interactive HTML for artifact canvas, only after tools work well in conversation

Cross-platform readiness: The MCP server is the product. MCPB is the first packaging for Claude Desktop. The same server works with any MCP-compatible client (ChatGPT, Gemini, VS Code, Goose). Persona adapts per platform.

Build the server cleanly enough that the packaging layer is swappable — anchor on the thin-wrapper-to-API model, not on MCP specifically.

Methodology Maintenance (Per CIO Observation)

Methodology-as-product is an asset only while it’s a living practice. The mechanisms that keep it alive:

• Trigger-based audit cadence — methodology audit within 2 weeks of each sprint gate closure (8-week max)
• CIO self-approval authority for Emerging patterns
• Cross-pollination review as standing intelligence infrastructure
• Per-sprint quality gate with CXO authority — each sprint needs a gate

These don’t appear in the feature roadmap but are required for the differentiator stack to remain credible.

Sprint Summary

Timeline (Inchworm, Not Calendar)

These are sequence statements, not deadlines. We are time lords. Each phase complete before the next begins.

April 2026

• M1 gate closure (Apr 11)
• M2 sprint begin (NOW)
• IAC presentation (Apr 17, Philadelphia)

Estimated forward sequence

• M2 (Conscious Floor + Action Handlers): 6-8 weeks (largest sprint, expansion expected)
• M3 (Artifact Persistence): 4-6 weeks
• M4 (Trust + Learning): 3-5 weeks
• M5 (Distribution + Polish): 4-6 weeks
• Beta via MCPB → v1.0

Change Log

• v15.0 (April 11, 2026): Major restructure post-M1 closure. Sprints reorganized around differentiator stack per Vision V2.3. 12 closures, 3 revisions, 10 new issues filed (#950-959), DIST integrated into M5. Leadership reviewed and endorsed (PPM, CXO, CIO, Architect). Build sequence and distribution strategy formalized.
• v14.3 (March 10, 2026): M0 marked complete, M1 cherry-picking documented (archived at docs/internal/planning/historical/roadmap-v14.3-2026-03-10.md)
• Earlier versions: see archived historical roadmaps