Product Dev OS

• You own the core product thinking, and AI fills in the gaps. By default, most AI coding tools want to own the problem, which may be fine for side projects. For real product work you need to own the fundamentals, and Product Dev OS will push you on this.
• Builds on your real stack and roadmap instead of 0 → 1. Most AI coding tools assume a ~single-shot, greenfield build, but 95% of us are building on existing platforms with context and flaws. Product Dev OS requires context on your company's strategy, customers, and constraints.
• It's doc-first (though you can build prototypes too). Larger features require documentation of edge cases and dependencies that don't fit neatly in a prototype. This system will help build eng-ready PRDs with all the gory details.
• Product decisions get sharpened and challenged instead of just agreed with. Work runs through purpose-built personas (Founder/CEO, Design, Tech Lead, QA, Support, GTM) with a high bar, so the system pushes back instead of blindly reinforcing.

Three inputs feed every artifact:

• PM POV: the AI will ask you your read on the problem, why it matters, your hypotheses, worries, and your conviction level. The AI builds on your judgment, not around it.
• Feature Context: drop in raw evidence like customer quotes, support tickets, sales feedback, competitive intel. Messy is fine.
• Company Context: our product vision, north star metrics, quarterly goals, design principles. Set once in System/company-context.md, reuse for every feature.

Conviction calibrates how the AI behaves:

Five modular artifact types: spin off a prototype from the brief as soon as it's ready; you don't need to finish the PRD first. Produce what the feature and its stage require.

Three inputs converge at the Artifact Hub, which routes to the right workflow. Each artifact runs Draft → Review → Revise; the AI simulates the right stakeholders and tags every piece of feedback P0/P1/P2.

flowchart TB
    subgraph inputs["Inputs"]
        A1[PM POV]
        A2[Feature Context]
        A3[Company Context]
    end

    B{Confident in solution?}
    B -->|High| B1[Sharpen]
    B -->|Mid| B2[Sharpen + challenge]
    B -->|Low| B3[Challenge]

    subgraph menu["Artifacts (draft with AI)"]
        M1[Product Brief]
        M2[Full PRD]
        M3[Prototype]
        M4[GTM 1-Pager]
        M5[External Docs]
    end

    subgraph swarm["Personas give feedback"]
        P1[CEO/Founder]
        P2[Tech Lead]
        P3[Design Lead]
        P4[QA · GTM · Support · Legal]
    end

    subgraph loop["Draft → Review → Iterate"]
        D[Draft]
        R[Review P0/P1/P2]
        I[Revise]
        D --> R --> I --> D
    end

    F[Final artifacts]

    inputs --> B
    B1 --> menu
    B2 --> menu
    B3 --> menu
    menu --> swarm
    swarm --> loop
    loop --> F
              

Review phase uses AI-simulated personas. Each focuses on a different slice (strategy, engineering, UX, compliance, etc.):