Stackmates

Each venture in the Mycelium follows the Business Ideas Template. This one is the platform the others run on.

Persuasion Loop Status​

1. Title & One-Line Description​

• Business name: Stackmates
• Domain: stackmates.io
• One-liner: Enterprise SaaS where humans and AI agents operate through the same business logic — a phygital architecture for coordinating biological and digital workers.

2. Problem & Why Now​

• What system is broken?

  • Business tools treat AI as an add-on, not a first-class operator alongside humans
  • Internal tools fragment across departments — no shared domain model
  • Agent orchestration is ad-hoc scripts, not structured coordination

• Who feels the pain most?

  • Operations teams managing workflows that cross human and AI boundaries
  • Founders building AI-native businesses who need structured coordination, not chatbot wrappers

• Why urgent now?

  • AI agents are production-ready but lack infrastructure to operate as team members
  • A2A (agent-to-agent) protocols are emerging — first movers define the coordination layer
  • The 2027 window — ~2 years before AI + data enters a self-reinforcing loop

• Why this approach?

  • Hexagonal architecture means the domain layer is pure — same business logic serves UI, API, and agent protocols
  • Commissioning model from factory engineering gives quantified maturity, not guesswork

3. System Map & Opportunity​

• Universe: AI-native business operations platforms

• Slot: The coordination layer where human intent meets agent execution

• Network:

  • Digital: AI agent teams, A2A protocol, shared domain model
  • Physical: DePIN integration path for real-world asset coordination

3.1 Open Protocol​

Stackmates is designed as public coordination infrastructure, not a closed productivity app.

This makes Stackmates usable as a build substrate for many ventures in the same digital mycelium.

Tight Five Biz Dev Idea​

Shared PRD Mycelium​

Stackmates is the shared PRD execution substrate for all mycelium windows.

4. Solution & Product Snapshot​

• Plain-language description: A platform where you define business operations once and both humans (via UI) and AI agents (via API) execute them through identical data paths. One domain model, multiple interfaces.

• Core job to be done: "When I have operations that need both human judgment and AI execution, help me coordinate them through one system instead of gluing tools together."

• Key differentiator: Phygital-first — AI agents are not bolted on, they share the same ports, use cases, and data paths as human users from day one.

DePIN to Robotics​

The long path is not "software only." It is software + sensors + robots coordinated through one business graph.

See Robotics Industry for the market layer and Games for incentive design.

5. Architecture​

Nx monorepo. 168K lines. 2,829 files. 50 Nx projects. 390 PRs merged. Solo architect and developer.

10-layer hexagonal architecture — inner layers never import outer:

The Orchestration System​

Designed from dairy factory commissioning — each data entity progresses through 10 discrete states from idea to fully commissioned (schema, migration, data, repository, server action, CRUD UI, ETL pipeline, A2A API, E2E tests, monitored).

• 5 AI agent teams (meta/orchestrator, UI, intelligence-engine, platform-engineering, marketing) operating in isolated git worktrees
• DB-native planning — all plan state lives in PostgreSQL, not files
• 15 plan templates standardise work the way Nx generators standardise code
• Cross-team composition — one plan assigns tasks to multiple teams
• Data footprint analysis algorithm scores all 208 tables by commissioning maturity

Agency Layer​

Client-safe library (runs in browser or server, zero infrastructure dependencies):

• 13 algorithms: data-footprint analysis, sales forecasting, RFP type detection, content strategy, decision-making, investment analysis
• 11 workcharts: sales RFP workflow, marketing strategy, marketing loyalty, slide deck generator
• A2A protocol: agent-to-agent intent API through /api/intents/ routes

Tech Stack​

TypeScript, Next.js 15 (App Router/RSC), Drizzle ORM, PostgreSQL (Supabase), Nx monorepo, Playwright, Vercel AI SDK, TailwindCSS v4, TanStack Table/Form, Clerk auth, Convex (real-time), Sui blockchain (Move)

6. Commissioning Status​

• Current stage: Active development
• 208 tables scored, average maturity 3.29/10
• Top 7 entities at 6.2–6.6 (need UI, ETL, A2A, E2E to reach 7.0+ commissioned threshold)
• 12 plans completed, 4 active, 96% task completion rate

7. Business Model Economics​

• Who pays: Teams and founders running AI-native operations
• Who earns: Platform takes margin on coordination, not compute
• Value capture: The domain model is the moat — switching cost is the business logic, not the UI

8. Team & Credibility​

• Founder: Matt Mischewski — CV
• Background: Mechanical engineering → factory commissioning → telecom routing → healthcare ETL → crypto/AI platform. Same pattern every time: unify siloed data, build reusable pipelines, surface decisions.
• Key relationships: Mycelium ventures share the platform

9. Key Risks​

10. The Ask​

• Current need: Reach 7.0+ commissioning on top 7 entities, ship first external user
• Not raising yet — proving the coordination layer works on own ventures first

10.1 Fair Trade Game​

Every workflow is a repeated game. If incentives are unfair, quality decays. If incentives align, trust compounds.

Goal: turn collaboration into measurable fair trades that produce both fulfillment and outcomes.

Connection to Mycelium​