Fixed-price removes billing uncertainty. You know the exact cost upfront and pay only when acceptance tests pass. If tests fail, I fix until they pass at no extra charge. The outcome is defined, not the hours, so no surprises.

Changes go through Swap or Add. Swap: Replace a feature with one of equal complexity at no charge (example: switch from web UI to Slack integration). Add: New feature becomes a separate milestone, priced upfront before work starts. This prevents scope creep while staying flexible. You still pay only when tests pass. Most changes are Swaps, which cost nothing.

Before starting, we co-write acceptance criteria - specific, testable requirements the prototype must meet. Example: "Answers 80%+ of questions correctly when tested with 100 sample queries" or "Processes 50 documents in under 2 minutes." I build automated tests for these criteria. You see test results (pass/fail) before final payment. If tests fail, I fix until they pass - included in the price. No guessing if it's done.

The prototype is production-quality code (clean, documented, tested) but scoped to prove ONE feature works. You won't rebuild - you'll extend it. Part of your deliverable is a detailed roadmap showing exactly what's needed to scale from prototype to full production: additional features, infrastructure setup, performance optimization, deployment, monitoring. Most clients use the Advanced tier or add Production Deployment as an add-on to go live immediately.

That's what acceptance criteria prevent. We define success metrics upfront - you write the tests. If the prototype fails those tests, I don't get paid until it passes. But if you mean "what if the approach itself is wrong?" - that's discovered in Week 1 during requirements workshop. If I identify a blocker (technical limitation, data issue, cost too high), I tell you immediately with alternatives. You never pay for work that can't succeed. Transparency over surprises.