Architecture
Generated 2026-08-07T15:25:56.947607Z
Operating Model
The factory runs on short feedback loops. Human attention is reserved for product direction, ambiguity, risk approval, and final acceptance — everything else is delegated to agents.
The founder sets product direction, priorities, and approval boundaries. Human attention is reserved for product direction, ambiguity, risk approval, and final acceptance.
Research agents inspect code, docs, market context, and risks before any changes.
Planning agents turn intent into scoped implementation plans with objectives, non-scope, constraints, test strategy, rollback, and human approval gates.
Implementation agents make the smallest correct change, run tests, and verify the result.
Review agents check correctness, regressions, security, tests, and docs. Each change includes a summary, rationale, files changed, verification, risks, and limitations.
Factory-improvement agents convert repeated friction into reusable systems, templates, prompts, tests, workflows, and docs.
Executive Organization
Agents are organized like an accountable company, not a pile of tools. Every role owns one decision; every agent reports to exactly one executive owner.
Chief Product Officer
Build the right products
Market research, customer feedback, product strategy, roadmap, PRDs, UX research, success metrics.
Chief Technology Officer
Build products correctly
Architecture, code quality, platform architecture, AI engineering, technical standards, security, performance.
Chief Operating Officer
Optimize the factory itself
CI/CD, deployment, quality gates, documentation, knowledge management, AI orchestration, workflow optimization.
Chief Revenue Officer
Maximize long-term customer value
Customer acquisition, pricing, subscription optimization, App Store optimization, retention, portfolio growth.
Chief Financial Officer
Allocate capital wisely
Cash flow, profitability, AI compute costs, API spend, ROI analysis, portfolio allocation.
Chief Knowledge Officer
Preserve and evolve institutional knowledge
Standards, prompt libraries, reusable templates, lessons learned, decision records, knowledge hygiene.
Agent Roster
Specialist agents handle research, planning, implementation, review, and factory improvement. A coordinating agent remains accountable for final integration and reporting.
Turns intent into scoped implementation plans with validation and approval gates.
Makes small, verified changes following the approved plan.
Checks correctness, regressions, security, test gaps, and maintainability.
Reviews screenshots, UI layouts, and rendered output for visual compliance.
Reviews generated SFX, ambience, and voice lines for quality and suitability.
Runs or analyzes local evals and validation results deterministically.
Reviews editorial and portfolio web-presence content for quality.
Cheap-tier agents for triage, summaries, retrospectives, and small docs edits.
Human Approval Gates
The factory is autonomous only inside approved boundaries. These categories always require explicit human approval before proceeding.
- Destructive operations — Data deletion, migration, or irreversible changes.
- Production deployment — Releasing changes to live production environments.
- Release submission — App Store review submission and TestFlight soak.
- Billing & payments — Subscriptions, store configuration, pricing changes.
- Secrets & credentials — API keys, tokens, certificates, permission changes.
- Dependency changes — Supply-chain or licensing impact additions.
- External services — Configuration of external service integrations.
- Legal & compliance — Privacy, policy, or compliance-sensitive changes.
- Architectural rewrites — Large-scale rewrites that materially alter product direction.
Trust Boundaries
What agents can do on their own vs. what requires human sign-off.
Autonomous
Read, search, validate, plan, implement within approved scope, run tests, generate docs, review code, create PRs (after review PASS), merge PRs (after review PASS).
Requires approval
Destructive operations, production deployment, release submission, billing, secrets, dependency changes, external services, legal/compliance, architectural rewrites, product direction changes.
Data Flow
From intent to shipped product to measured improvement — a closed loop.