Guides
Build lab
Optional deep dive: build one app end to end and learn the whole Vercel AI stack, updated for AI SDK v7.
Build one app end to end and learn the whole Vercel AI stack. You start by typing a sentence into v0 and finish with a multi-tenant, agentic, evaluated application running in production. Every level builds on the same app, every step is copy-pasteable, and every level ends with a check so you know it worked before moving on.
Heads up — this is an optional deep dive This lab goes far beyond what you need for the two-week hackathon. The event is part-time, and you only need to ship ~25% of one idea — no runtime AI is required. Treat this as a "keep going after the event" track for anyone who wants the full picture. Skim the parts you need; ignore the rest.
Updated for AI SDK v7 The code here targets AI SDK v7, the current release. If you paste snippets from older blog posts you will hit renamed APIs — the callouts flag the ones that moved.
Who this is for
Engineers who can already ship a Next.js app and want to know where the leverage actually is. Levels 0–2 are deliberately brisk — skim them if you already deploy on Vercel. The compounding starts at level 3, and levels 8–14 are where most teams leave quality on the table. Every level ends with primary-source references, because the ecosystem moved fast enough that half the blog posts you'll find are wrong.
Roughly 9 hours total: 4 beginner levels, 4 intermediate, 7 advanced.
Chapters
Six chapters, in order. Each stands on its own, but they build on the same app.
1. Foundations
Levels 0–3Set up the tooling, generate your first app from a prompt, ship it through a PR, and build the context layer every later step depends on.
Beginner~90 min2. Data & access
Levels 4–5Real Postgres with shared Zod contracts, then auth, organisations, and server-side authorisation you can trust.
Intermediate~70 min3. AI features
Levels 6–7Stream your first model response through AI Gateway, then turn text into typed, validated, stored data.
Intermediate~65 min4. Agents & tools
Levels 8–11Build a real agent loop, gate risky actions behind human approval, run model-written code in a sandbox, and make runs durable.
Advanced~2.5 hrs5. Production
Levels 12–14Runtimes and caching, hardening and cost observability, and an eval suite that makes the whole thing maintainable.
Advanced~2 hrs6. Cheat sheet
ReferenceCommands, the API changes that break old tutorials, the habits that mattered, and a primary-source reference index.
Reference5 min