Where the code runs
Laptop, browser, always-on server, function that wakes up, edge, phone. Frameworks are packaged answers to this question. Start here.
A worksheet, not a chatbot
Pick an ambition tier before you pick a stack.
Amateur vibe coders keep asking language models for a tech stack and getting Next.js, or Python, depending on the weather. Six questions about your actual idea work better — they teach the physics frameworks are hiding.
The output is a tier, a default shape, a list of things you do not need, napkin math, and three pages to read. Not a shopping list.
The tiers
These are levels of architectural ambition, not skill. Most ideas should stay on 0–2.
| Tier | What it is | Typical cost | Default shape |
|---|---|---|---|
| 0 · Local | Runs on your machine | $0 | HTML file, script, notebook |
| 1 · Static + glue | Files on a CDN, maybe one worker | $0–5/mo | Hand HTML or an SSG, Pages |
| 2 · One process | A real server and one database | $5–20/mo | One app + SQLite or Postgres |
| 3 · A product | Users, email, uploads, payments, jobs | $20–100/mo until it works | Same app, still one DB |
| 4 · Scale theater | Multi-region, k8s, microservices | A second job | You do not start here |
Static site, actually
Astro is an LLM default that is also simple. HTMX is an LLM default that usually is not the right primitive for a brochure. The difference is the point.