Solo Founder Path
Who it's for
Founders and indie hackers who are the product, architecture, and review board. Agents multiply throughput; they also multiply wrong product decisions if you skip specs.
Prerequisites
- Ability to ship a vertical slice (auth → core loop → deploy) yourself if the agent fails
- One product idea with a real user in mind
- Willingness to cut scope ruthlessly
Ordered modules
- Mental Models — intent quality bounds output quality
- Prompts vs Specs — chat is not a product spec
- Idea to Spec — force problem, users, non-goals
- New Project — scaffold with memory files from day one
- MVP Build — ship the slice, not the platform
- CLAUDE.md & AGENTS.md — cheap insurance
- Testing AI Code — you have no review buddy; tests are the buddy
- Prototype to Production — what breaks after the demo
- SaaS MVP case study — realistic path and mistakes
- Recovery — when the agent paints you into a corner
Exercises
- Fill a new project brief and MVP spec; cut until a two-week build is plausible.
- Create the repo with CLAUDE.md + ARCHITECTURE.md stubs before feature code.
- Build only the in-scope MVP acceptance criteria; keep a dated list of agent-suggested features you refused.
- Run the deployment checklist before first real users.
Capstone
Ship an MVP to real users (or paying design partners) with: filled MVP spec, agent-built implementation you can explain, basic verification, and a one-page post-ship note of what the agent invented that you deleted.
Expected outcome
You can go from idea → deployed MVP without a team, while keeping architecture and scope under your control — not the model's.