When making gets cheaper, taste matters more (Simon Willison)

Simon Willison published a quietly remarkable piece of evidence in late October: four browser games he built across a few weeks, none of them representing more than four hours of his own time, though he is careful to note the model spent many more hours churning away by itself. The prompt he credits with transforming one of them, a real-time strategy game, was "Make it a LOT more fun."

The detail worth attention is where his effort went. The starting ideas were his, and he recognised when a prototype was fun rather than merely competent and decided what to ask for next, while the model produced the code, the art, the music and much of the design. Of one game he writes that his creative input was "pretty much just 'it's a fantasy village and there's a troll asleep in a cave'". His conclusion is that the new challenge is coming up with concepts that would not exist without AI assistance, and that the interesting question is what happens when someone with better taste actually puts the work in.

Care is needed with how far this travels. These are personal games built for delight at low stakes, not delivered software carrying customers, regulators and a maintenance tail, and the cost of a prototype is not the cost of a product.

However, read as an operating-model observation rather than a finding, his account points somewhere useful. If producing a competent version of something keeps getting cheaper, then more of the value moves to selection: specifying what good looks like before the market confirms it, and telling the third draft from the fourth. That is my reading of his evidence rather than his claim, and it raises a genuine design question. Where does taste sit in an organisation, and is it in the room when the prompts are written, or only when the launch is approved?

#AI #Product #Innovation

Sources:
- Simon Willison, "The October inflection point for vibe game development", 28 October 2026: https://lnkd.in/gBHMS3Hi