These three words get blurred constantly. They shouldn't be — they answer different questions, take different amounts of time, and cost different amounts of money. Pick the wrong one and you spend 12 weeks answering a question you could've answered in 12 hours.
Proof of Concept (PoC): "Is this technically possible?"
A PoC answers a feasibility question. Can the API actually do what I think? Can this model hit the latency I need? PoCs are not for users. They're for you. Build only when there's real technical risk — most app ideas in 2026 have zero technical risk.
Prototype: "Does this feel right?"
A prototype is a clickable mockup. Figma, a design tool, even slides. It answers UX questions: do users understand the flow? Is the value clear in 5 seconds? Show it to 10 target users before writing code. Prototypes save weeks.
MVP: "Will people pay?"
An MVP is a real, working product narrow enough to test the commercial hypothesis. It should do one thing end-to-end well enough that someone hands you money. Not five things mediocre. One.
The decision tree
- Real technical unknowns? → PoC first (1–3 days).
- Flow or value-prop unclear? → Prototype (3–5 days).
- Demand unproven? → 48-hour demand test, not an MVP.
- Demand proven? → MVP, scoped to one workflow (2–3 weeks max).
The trap most founders fall into
They call a 12-feature, 3-month build an "MVP." It isn't. That's a product. A real MVP is embarrassingly small — see The Vibe-Coding Trap for why AI tools make this trap worse, not better.
The honest sequence
Validation → Prototype → MVP → Product. Skip steps and you'll either build the wrong thing or build it for nobody. The good news: validation and prototyping together cost a long weekend.