The 25-Minute Build
You have 25 minutes and an AI coding agent. Pick the scope you would actually build. This playground gives you a fixed menu of tasks — some core to a demoable result, some pure scope creep — and scores whether your chosen scope both fits the clock and ships something you can show. It turns the slogan "AI makes you 10x faster" into a falsifiable question: at the buzzer there is either a working, tested artifact, or there isn't. Runs entirely in your browser (0 uploads, works offline).
How to use it
- Read the eight tasks — each shows its estimated minutes and whether it is core (ships) or scope creep.
- Check the tasks you would build in the window, or load a preset — Tight scope or Scope creep.
- Watch the clock: the meter fills toward 25 minutes, and the badge tells you whether your scope fits and ships, busts the clock, or ships nothing demoable.
- Open the X-ray panel to see exactly how many of your minutes went to polish that made nothing more demoable — then cut it.
What this clears up (the fundamentals)
- Speed is scope, not typing — the reason a build fits in 25 minutes is a ruthless scope, not a faster keyboard. The agent only helps once you have decided the one thing to ship.
- "Shippable" has a definition — it means demoable: a working, tested happy path. Login, a landing page, and a dark-mode toggle can all be real work and still ship nothing you can show at the buzzer.
- Scope creep is invisible until you count it — polish feels like progress. Summing the minutes it costs makes the trade-off honest: every creep task is a core task you did not build.
- The clock makes the claim testable — "10x faster" is either proven by a visible artifact inside the window or it is not. A time box converts an assertion into evidence.
Where it's used
A FREE, top-of-funnel check for AI-assisted engineering — the fastest way to feel why scoped delegation and time-boxed supervision beat open-ended prompting. It is a miatz build-lab concept playable: play the scoping game here, then learn to run the real graded 25-minute sprint — live sandbox, AI agent, pass/fail at the buzzer — in the AI Coding Dojo.
FAQ
Why 25 minutes?
It is long enough to build one genuinely completable thing and short enough that scope discipline is the whole game. The number is the point: a hard time box forces you to name the single demoable result before you start.
What counts as "shippable"?
A demoable, working happy path — ideally with one smoke test proving it runs. Tasks marked scope creep (auth, a landing page, a theme toggle, logo polish) can be perfectly good work, but they do not make your demo exist, so they do not count toward shipping.
Does the tool call an AI model?
No. This teaching version is a deterministic scoping game — it sums minutes and grades your scope entirely on-device. The real graded sprint (BYOK, live sandbox, an actual coding agent) lives in the paid AI Coding Dojo; here you practise the decision that makes those sprints winnable.
Is anything uploaded?
No. Your task selections are scored entirely in your browser — nothing is transmitted, stored, or logged. Turn off your Wi-Fi and it still works.
Limits
A teaching model with eight fixed tasks and hand-set minute estimates — it sharpens the scoping instinct, it does not estimate your real project. Actual build time depends on the brief, the stack, and how much the agent gets right on the first try. The lesson it enforces (cut to one demoable thing and time-box it) is exactly the real practice.
Related
Part of the Demystify Playgrounds. Explore the rest from the Playgrounds home.
Bookmark this page (Ctrl+D, or ⌘D on Mac) — it works offline the next time you need it.