Three-Prototype Ideation
Since prototypes are now free, build a feature three ways and let real use pick the winner.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 3
- Confidence
- 90%
A product-design method that exploits the collapse in the cost of building working prototypes. Instead of speccing one option and betting on it, you prototype a feature several different ways in the time it used to take to build one, then experiment with all of them to find which actually works. It shifts the bottleneck from building to proving, and reserves human judgment for evaluation rather than production.
Origin
Described by Simon Willison from his own practice; he notes his lifelong superpower was rapid prototyping and that AI has democratized it, forcing the value to move to knowing when and how to prototype.
Core principles
- 01Your initial ideas are always wrong — what matters is proving them
- 02A UI prototype is effectively free now: ChatGPT and Claude will build a convincing UI for anything you describe
- 03Writing code is cheap, so prototyping is almost free — this is the transformational step in ideation
- 04Product designers not vibe-coding prototypes are missing the most powerful boost available
How to run it
- 1
Take a feature idea and prototype it three ways
For any feature you want to design, have AI build roughly three different working versions of how it could function. This takes very little time.
Pro tip Treat the UI prototype as free — describe it and let the model generate it rather than agonizing over specs.
- 2
Experiment with each to see which you like
Play with the variants yourself, trying them and forming an early read on which directions feel right.
- 3
Prove the winner with real humans, not simulated ones
To decide between surviving options, fall back on old-fashioned usability testing — put a real person on a screen-share and watch them use it.
Watch out Do not trust an AI pretending to click around your prototype; simulated users give far weaker signal than a real human being.
In the wild
Willison connects this to naming expert David Placek, who briefs three teams on the same product from different metaphors — e.g. name it as an AI IDE, as a boat, as a spaceship. The best names come from the non-literal directions that share the product's benefits.
→ Divergent framings surface stronger options than a single on-the-nose attempt — the same logic as building a feature three ways.
Common mistakes
Betting on your first idea
Initial ideas are the beginning of an idea, not the winning one; skipping the multi-prototype exploration wastes the now-cheap ability to test alternatives.
Using AI to simulate your usability test
Having ChatGPT pretend to be a user clicking around gives non-credible results — the evaluation step still needs a real human.
Is it for you?
Best for
Product designers and PMs deciding how a new feature should work
Not ideal for
Well-understood problems with a single obvious implementation, where three variants is wasted effort
From the transcript
“any sort of feature that I want to design, I'll often prototype like three different ways it could work”
“a a a a UI prototype is free now.”
“I don't think you're going to get as good results from ChatGPT pretending to click around on your prototype than you would from an actual…”
From the episode
An AI state of the union: We’ve passed the inflection point, dark factories are coming, and automation timelines
Simon Willison