Skip to main content

Part 9: The Solo / Indie AI Workflow

For the one-person builder shipping AI side projects, demos, and indie apps in 2026.

In one line: Frontier API by default, free tiers everywhere, ship in days. Skip every "real engineering" practice that doesn't earn its keep at your scale.

In plain English

Who this chapter is for: You, alone, on a laptop, with a Claude or OpenAI API key. No team. No infra. No users yet. You want to ship an AI thing this weekend that real strangers can use on Monday.

The whole philosophy of solo AI in 2026: Don't fine-tune. Don't self-host a model. Don't build an eval platform. Don't write a custom RAG framework. The frontier API is good enough, the managed-everything stack is good enough, and your time is the only resource that doesn't scale. Spend it on the product.

What "shipping AI" actually means here: Someone you've never met opens a URL, types something, gets an LLM-generated response, and the bill at the end of the month doesn't make you cry. That's the whole game at this stage.

The 2026 solo AI stack at a glance:

  • Model: Claude Sonnet 4.5 or GPT-5 mini, called via the official SDK
  • App framework: Next.js + Vercel AI SDK (TS) or FastAPI + Modal (Python)
  • DB + auth: Supabase (Postgres + pgvector + auth in one)
  • Hosting: Vercel free tier or Modal pay-per-invocation
  • Observability: Langfuse free tier
  • Payments (when you need it): Stripe Checkout or Polar
  • Total monthly cost at hobby scale: $0–$50

Mental model: A solo AI project is like running a food truck, not opening a restaurant. You don't need a sommelier, a maître d', a 60-page menu, or a Michelin star. You need one dish that's good enough that strangers will pay for it, served from a vehicle you can drive yourself.

If you only remember one thing: The best solo AI stack is the one that gets you from "idea" to "live URL with auth and a cost cap" by Sunday night. Everything else is procrastination dressed up as engineering.

How this chapter is organized​

Each page focuses on one slice of the solo AI workflow. Read them in order the first time; jump back to any one when you hit that phase on a real project.

The pages are deliberately short and opinionated. There's a whole industry telling you to fine-tune your own 7B model on synthetic data. Ignore it. You're building a thing one person uses on a Friday afternoon.

Pages in this chapter​

  1. The Solo AI Builder Mindset — Why "real ML engineering" advice usually destroys solo AI projects.
  2. What Kinds of AI Side Projects Actually Work Solo — Single-prompt tools, niche RAG, narrow agents, AI-augmented existing tools.
  3. Planning a Solo AI Project — The pre-mortem, the one-pager spec, and eval criteria before you write code.
  4. Stack Selection — The two default 2026 stacks (TypeScript and Python) and when to pick which.
  5. Environment Setup — Dotfiles, env vars, secrets, and the seven free-tier accounts to open on day one.
  6. The Development Loop — Prompt iteration in a Jupyter/REPL, fast eval scripts, the "edit prompt, re-run eval, commit" rhythm.
  7. Auth — Clerk or Supabase auth, per-user rate limits, and why anonymous LLM endpoints get abused within hours.
  8. Payments — Stripe Checkout, Polar, Lemon Squeezy, and usage-based pricing for AI tools.
  9. Deployment — Vercel, Modal, Render, Fly; environment promotion; preview deploys.
  10. Observability — Langfuse free tier, the minimum useful traces, and a cost dashboard that pages you.
  11. Launching — Distribution channels for AI tools in 2026 and the launch-tweet template.
  12. Maintenance — Model deprecations, provider price changes, and eval drift on a 15-minute weekly cadence.
  13. Realistic Time Investment — Weekend MVP, month to 100 users, three months to $100 MRR.
  14. Common Pitfalls — The ten things that kill solo AI projects.
  15. Starter Templates — Streaming chat, RAG over docs, structured-output classifier, simple agent.
  16. Sample Project: AI Meeting-Notes Summarizer — End-to-end walkthrough with code, costs, deploy, and first users.
  17. Graduating Beyond Solo — When to bring on a co-founder, when to convert to a real company, when to keep it indie.
  18. Chapter 9 Checkpoint — Self-test before moving on.

→ Start with The Solo AI Builder Mindset.

When you finish all 18 pages, move on to Chapter 10: Startup AI Team.