A PM-first playbook for choosing the right AI build tool by job-to-be-done.
Not tool tutorials. Cross-tool decision frameworks for PMs and founders making real execution tradeoffs.
Maintained by Sumit Soni · LinkedIn
- Product Managers who ship, not just spec — you own the build decisions, not just the roadmap
- Non-technical founders who want to move from prototype to production without a full engineering team
- NOT engineers or ML practitioners (there are better resources for you)
When to use Claude Code vs Codex vs Gemini CLI vs v0 vs Bolt — by product stage and job-to-be-done.
| Tool | Best for | Avoid when |
|---|---|---|
| Claude Code | Structured SDLC, repo-level context, production builds | Quick throwaway prototypes |
| OpenAI Codex | Agentic experiments, broad code exploration | Precision edits to existing codebases |
| Gemini CLI | Large-context analysis, research, multimodal | Speed-critical coding tasks |
| v0.dev | UI-first prototyping, React components | Backend logic or data pipelines |
| Bolt.new | Full-stack browser prototyping, fast demos | Production-hardened deployments |
- Discovery to prototype — PRD → clickable prototype in under 2 hours using v0 + Claude Code
- Feedback to fixes — User feedback → prioritized fixes shipped in 1 day using Codex + triage templates
- Codebase audit to decision memo — Large-codebase audit → PM decision memo using Gemini CLI
- Prototype to production — Prototype → production-hardening checklist using Claude Code + promptfoo eval gates
- Cost spike to model routing — Cost spike diagnosis → model-routing redesign using telemetry + tiered fallback strategy
- Debugging to root cause — Claude-generated code breaks; diagnose and fix it without guessing
- Spec to sprint — Feature spec to sprint-ready tasks with acceptance criteria an engineer can act on
- Founder path — no engineering team, shipping first AI product
- PM in startup path — small team, fast iteration cycles
- PM in larger org path — stakeholder management, eval discipline, rollback plans
- Not a beginner tool tutorial catalog (see Carl Vellotti's ccforpms.com for that)
- Not a curated resource list (see awesome-ai-pm for that)
- Not an engineering reference
Most vibe-coding repos optimize for engineers. They skip:
- PM artifacts — problem framing, acceptance criteria, launch readiness
- Tool-choice heuristics by product phase
- Evaluation discipline for PM-owned quality decisions
- Stakeholder narrative layer — how PMs communicate risks and tradeoffs
This repo fills those gaps.
PRs welcome if you bring workflow evidence, not just opinions.
Requirements:
- Include a real scenario with context (what stage, what team size, what constraints)
- Include failure mode notes — what breaks and how you caught it
- No generic "use AI to move faster" advice
- pm-prompts — Prompt library for every PM workflow
- awesome-ai-pm — Curated AI PM resources
- purepaste — Live product built with this stack
Star this repo if you find it useful. Playbook content shipping weekly.