Week 1: the scoping sprint that saves the project
Most MVP failures are scoping failures wearing engineering costumes. Week one produces: the wedge statement (the one hypothesis this MVP must test), user-flow maps for the two or three core journeys, the cut list (everything explicitly not being built, signed by both sides), stack decisions, and a fixed-price quote with milestone gates.
The cut list is the artifact clients resist and later thank us for — an MVP that tests the wedge in four weeks beats a 'complete product' arriving in four months, after the market moved.
Weeks 2–3: build sprints without mystery
Stack defaults (adjusted per project): Next.js on Vercel, Supabase for auth/database/storage, Claude or Gemini APIs where AI features are core, Stripe if payments matter day one. Boring, battle-tested, cheap to run and easy to hire for later — the stack a startup should inherit, not a consultancy's pet framework.
Process mechanics: twice-weekly demo calls on the live staging URL, a shared task board the client can read, and mid-build scope-change rules (anything new goes on the v2 list unless it displaces something of equal size). Demos on real deployments kill the two classic agency pathologies: invisible progress and launch-day surprises.
- Next.js + Supabase + Claude/Gemini + Stripe as defaults
- Twice-weekly demos on live staging
- Scope changes trade, never add
- Client-readable task board throughout
Week 4: hardening, launch and what happens after
The unglamorous week that separates demos from products: auth edge cases, error states, mobile passes, analytics events on the metrics the wedge test needs, SEO basics, and production deploy with monitoring. Handover includes the repo, infrastructure access, documentation and a Loom walkthrough — you own everything, no hostage code.
Pricing honestly: simple validation builds land $10k–18k, standard SaaS MVPs $18k–30k, AI-native or marketplace complexity $30k–40k. Market rates for equivalent scope run $7k–25k+ at 14–30 days from competent shops — we're mid-market on price and opinionated on process, which is the actual product.
Questions we hear about this
The trade rule: new scope enters only by displacing equal scope, otherwise it's first on the v2 list. This isn't rigidity — it's what keeps 30 days true. The scoping week exists precisely so mid-build discoveries are rare and small.
When the wedge test allows it, we'll recommend it honestly — a Softr/Bubble build at $3k–8k sometimes beats custom code for pure validation. We build custom when the product is the moat, AI behavior is core, or the no-code ceiling would arrive within months.
You do, from day one: repo in your org, infrastructure on your accounts, keys in your vault. Agencies that hold code hostage are financing their retention problem with your equity. Our retention mechanism is being good to work with.
Three options, no lock-in: you take it fully in-house (docs and walkthrough included), a maintenance retainer for fixes and small features, or a v2 sprint when the wedge data justifies it. The wedge metrics from week four tell you which — that was the point of instrumenting them.