CONCEPT

Minimum Viable Product

The maximum learning instrument, not the minimum product — a discipline of epistemic restraint now stripped of its economic rationale by AI and revealed in its essential form.

The Minimum Viable Product was never a product strategy but a learning strategy: the smallest artifact that could test a hypothesis about the customer. Its minimalism was not a concession to economic constraint but a requirement of epistemic hygiene — a product containing ten features makes it impossible to determine which drove customer response, regardless of how cheaply the ten features were built. The AI revolution has removed the financial pressure that previously enforced restraint, exposing the MVP's essential logic: restraint is not about cost management but about protecting the signal from confounding variables. The MVP in the AI age must be reconceived not as the minimum product but as the maximum learning instrument — its constraint determined not by how much can be built but by how much can be learned from what is built.

In the AI Story

Hedcut illustration for Minimum Viable Product — Minimum Viable Product

The naive AI-era response to the MVP is celebration: if building is free, build more; if the full product costs the same as the prototype, skip the prototype. This response misunderstands what the MVP was designed to accomplish. The cost of the experiment was never primarily the cost of building. It was the cost of ambiguity, and ambiguity costs exactly as much in a world of free building as it did in a world of expensive building.

The Orange Pill's account of Napster Station illustrates the point. AI-assisted development compressed the timeline from quarters to weeks, but the learning timeline was not compressed. The kiosk still had to confront real users in a real environment, process their reactions, accommodate their unexpected behaviors, and reveal through the friction of actual use which assumptions were correct. The Build phase compressed by an order of magnitude; the Learn phase operated at its own pace, dictated by the complexity of human behavior.

The Boardy AI analysis of Lean Startup in 2025 documented founders building impressive apps with AI assistance and then realizing they had isolated themselves from genuine user feedback. 'You know who never challenged my idea? ChatGPT! It always sounds confident and supportive.' The AI had made building so frictionless that the founder never encountered the resistance that would have forced him to question his assumptions before the market delivered its verdict. The tool whispers: you can build everything. The methodology responds: you should build the one thing that will tell you what you need to know.

The standard of viability has also shifted upward. Tools can produce artifacts of remarkable polish in short timeframes, and the customer's baseline expectation has risen accordingly. A product that would have been acceptably minimal two years ago may now feel unfinished, and the customer's response to perceived incompleteness contaminates the data. The paradox: the MVP in the AI age may need to be more polished than the pre-AI MVP, even though polishing is no longer the hard part. The hard part is ensuring polish does not obscure the hypothesis.

Origin

Ries introduced the MVP concept formally in The Lean Startup (2011), drawing on Steve Blank's customer development methodology and his own experiences at IMVU. The term became one of the most widely adopted — and most widely misunderstood — concepts in contemporary entrepreneurship, frequently reduced to 'ship something fast' while losing the underlying epistemological discipline.

The concept's evolution in the AI age has been driven by practitioner debate rather than formal revision. Analyses like Boardy AI's 2025 examination document founders making 'micro-pivots' — continuous adjustments that blur the boundary between iteration and pivoting — suggesting the MVP is evolving from fixed artifact to continuously reshaped probe.

Key Ideas

Minimalism is epistemic, not economic. The MVP is minimum because testing one hypothesis at a time is the condition for drawing valid conclusions, regardless of how cheap building becomes.

The tool removed the filter. Implementation cost previously screened out ideas that could not survive the rigors of building. The filter must now be supplied by the builder's judgment.

Premature elaboration is the new pathology. Where pre-AI founders built too much before testing, AI-assisted founders test too frequently and learn too shallowly — each interaction triggering a new build before the previous cycle's learning has been absorbed.

The viability threshold has risen. Customer expectations of polish have increased with tool capabilities; the MVP must be more finished than ever without obscuring the hypothesis.

The micro-pivot is double-edged. Continuous adjustment is powerful when driven by accumulated learning and pathological when driven by inability to sit with ambiguous data.

Debates & Critiques

Some practitioners argue the MVP is obsolete in the AI age — that continuous reshaping of the product in response to real-time data replaces the discrete artifact the MVP represented. Ries's response, implicit in Solveit's architecture, is that the MVP's essential function — forcing clarity about what is being tested — cannot be performed by continuous reshaping, because continuous reshaping typically prevents the hypothesis from being cleanly formulated at all.

Appears in the Orange Pill Cycle

Eric Ries — On AI