Season to Taste: How Are SOTA Models Changing Creativity

This week ChatGPT dropped under half the market for the first time. While Gemini climbed to a quarter, and Claude took a slice.

Gemini's everywhere because it's already there - baked into the phone and the work apps by default. That's a distribution win, and distribution wins tend to end as the default nobody actively chose. Sometimes that default eventually earns the choice (Excel, search). Often it just sits there, unloved. Ask Siri.

Claude's numbers are interesting for a different reason. Reported paid-conversion sits highest in the field, and anecdotally the people paying skew senior - the ones who sign things and run rooms. Treat that as a vibe, not gospel; because from my perspective the data is still thin. But if it holds, it's the market quietly admitting it values judgment over the quantity of output. And the gap between what's cheap and infinite and what someone will still pay for is only going to widen.

Which is why I think the "AI slop" panic has it backwards.

Everyone treats the sameness as a bug - something better prompts will fix. It isn't a bug. It's the new ground floor. When competence becomes free and endless, competence stops being worth anything, and value floats up to whatever's still rare.

Right now that's taste. Knowing the difference between something good and something trash. 

Those insane new job titles - "Creative Guru," "Agent Whisperer" - sound stupid but point at something real. Because labour got faster and cheaper, the job is no longer making the thing. It's deciding what's allowed out the door. Fewer hands building means there should be more people standing at the exit doing QC.

That's the whole shift - When making something drops to roughly free and roughly identical, the thing itself stops being worth anything. So value drifts upward - into judgment, into brand, into the weirdly rare skill you’re finally sharing with the office.

Hito Steyerl once described the poor image as a thing that travels by degrading - losing resolution as it gains circulation, a lumpen proletariat of the visual. The image of 2026 inverts her: it gains resolution as it loses authorship. Sharper than anything a camera ever saw, and made by no one.

The businesses that win won't be the ones using the most AI. They'll be the ones who figure out soonest what makes them unique, what their authorship looks like.

I don't know if taste will survive the next round of . And I’m asking what's irreplaceable above taste?