“Build me a great website.”
The goal, audience, proof, operating systems, and acceptance criteria are all left for the model to guess.
Fast start · unstable finishPrompt Stack Simulator · Guided 0→1
One prompt asks.
A stack builds.
Watch a vague idea gain an outcome, audience, capabilities, a builder handoff, and a definition of done.
Run the simulator No login · no model call · no invented executionChoose an outcome
Three ways to go from zero to one.
The stack stays consistent. The evidence and operating jobs change with the outcome.
Live guided compile
Turn a prototype into something people can understand and try.
Build readiness37%
Visible impact
See what changes from step 1 to 3.
1/3 foundational decisions resolved
State what should exist when the work is done.
Capability modulesChoose the operating jobs
3/3 includedOutcome receipt
Complete the stack to reveal the handoff.
# Template Mode Guided Build Specification
> Deterministic educational output. No external AI or service has run.
## 1. Outcome contract
Build: A credible experiment page with a working demo path and honest limitations.
Starting idea: Launch an AI experiment that explains the problem, the method, and where the model fails.
## 2. Audience and evidence
Primary audience: Early users, collaborators, and technical reviewers
Evidence to make inspectable: Runnable demo, method notes, limitations, and a clear next action
## 3. Required capabilities
- Live demo: Give the visitor a safe way to try the core behavior.
- Proof ledger: Connect claims to methods, evidence, and limitations.
- Interest capture: Turn qualified curiosity into an owned follow-up.
## 4. Builder adapter
Produce a responsive, production-ready implementation. Keep canonical data and decision logic owned. Treat external models and services as replaceable infrastructure. Mark unknown facts as Needs input.
## 5. Acceptance and QA
Verify the primary journey end to end, keyboard access, 390px mobile layout, reduced motion, media fallbacks, truthful claims, explicit failure states, and no unsupported external action.
## 6. Expected artifact
AI Experiment Launch: A credible experiment page with a working demo path and honest limitations.Why the stack works
A one-shot prompt compresses decisions.
A stack makes them inspectable.
Outcome + context + capabilities + adapter + QA.
Each decision has an owner, a visible input, and a bounded contribution to the final artifact.
Slower framing · reusable result0Ambiguous idea01Base template02Personalization03Capability modules04Builder adapter05Acceptance and QA1Shippable outcome