TL;DR
New is the smallest part
Great products rarely start from blank-slate invention. They start with behavior already proven in the market, add one improvement strong enough to make people switch, then test one new idea without pretending it is likely to work.
Mark Pincus explains the pattern as Proven, Better, New: copy the job people already want done, make one part clearly better, then add one new bet you assume may fail. The useful discipline is not copying for its own sake. It is sequencing risk so the team knows what it is testing.
Attribution and thanks: this piece builds on Mark Pincus's conversation with Lenny Rachitsky on Lenny's Podcast about product invention, Zynga, retention, AI, and his book, Life at the Speed of Play.
Copy the job, not the surface
The proven part is not a license to clone lazily. It is a demand to understand why something works.
Pincus calls it getting your PhD in proven. Study the mechanics, the user habit, the distribution channel, and the emotional reward before you start decorating the idea with originality.
School trains people to treat copying as cheating. Product punishes that reflex. The market does not reward a founder for being original in the eyes of other founders. It rewards a product for being useful in the eyes of the customer.
Better has to be undeniable
The better part is not a positioning line. It has to be visible to the person already using the old thing.
Pincus sets a high bar: 10 out of 10 existing users should understand why they would switch. That is different from saying the product is fresher, more modern, or more founder-pleasing.
This is where ambition gets smaller and sharper. One better reason beats ten vague differentiators.
New is where hope sneaks in
The new part feels like the founder's real contribution, which is exactly why it is dangerous.
Pincus separates instinct from idea. The instinct can be right while the specific idea is wrong. So keep the instinct and cheapen the idea.
That is the operating lesson. Do not protect the first idea because it came from a good instinct. Protect the instinct by testing more ideas around it.
AI should make failure cheaper
This is where AI changes the product loop.
Pincus describes AI as a failure machine: a way to test many versions of an idea fast, not a way to spend three months polishing one bet. A failure machine is a system that returns evidence cheaply.
That is the useful version of AI in product work. Not "build the whole thing for me." More like: make the next bad version cheap enough that we can learn from it today.
Distribution is part of the product
The framework also forces one unglamorous question early: how will this reach people?
Pincus argues that app discovery is effectively broken, so distribution cannot be a post-launch hope. If the product needs a channel, a loop, a community, or a social mechanic to work, that belongs in the design from day one.
Proven, Better, New is not a slogan for copying. It is a way to remove false bravery from product work. Start with what people already do. Make one thing clearly better. Test one new bet. Then be honest about what worked.
