Prototyping in Production Is Underrated
The real product teaches you things a prototype cannot, as long as you ship with care.
I like prototypes. I also think some of the best prototyping happens inside the real product.
A prototype is great for exploring direction. It helps a team see the shape of an idea before committing to implementation. But prototypes can also lie. They do not have messy data, slow connections, permissions, empty states, edge cases, legacy constraints, or real users arriving with real stress.
Production reveals those things quickly.
That does not mean shipping careless experiments. It means designing small, reversible changes that can learn from reality. A new onboarding step behind a flag. A clearer empty state. A guided summary in a dashboard. A new explanation pattern for recommendations. A small admin tool used by one internal team before it becomes official.
The product becomes the research environment, but only if the team respects the user. Measure the change. Watch behavior. Provide escape routes. Keep the blast radius small. Do not turn users into unwilling QA.
The real product teaches faster because it has gravity.
AI makes this more interesting because the cost of building small versions has dropped. A designer can prototype closer to implementation. A builder can test more variations. A team can move from idea to real usage without waiting for every detail to become perfect.
But the discipline matters more, not less. Fast production prototyping needs clear intent, instrumentation, rollback, and a strong definition of what would make the experiment harmful.
When done well, prototyping in production is not reckless. It is honest. It admits that the real product is where the truth lives.
— Victor