Apple Announced a Brand New Siri. It Announced the Same Siri in 2024.
Apple just announced a brand new Siri.
It announced the same Siri in 2024.
That announcement cost them a reported $250 million.
A demo that shipped to the ads but never to the phones
In 2024, Apple demoed the new Siri on stage. Then it did something bolder than the demo: it ran the unshipped feature in iPhone 16 advertising. People bought the phone partly for a capability that didn't exist yet. The feature was real enough to sell and not real enough to use.
It never shipped. Last night, Apple demoed it again. Still not shipped. "This fall," in beta, and not in Europe.
Here's the part everyone scrolls past.
Building it was never the hard part
The new Siri reportedly runs on Google's Gemini, rented for something like a billion dollars a year. Set aside, for a moment, what it means that Apple is renting the brain of its flagship assistant from its biggest rival. Focus on what it tells you: the capability was available. It could be bought. The engineering came down to a procurement decision.
If the hard part were building it, a company with Apple's resources and a checkbook pointed at Google would have shipped it. They didn't. So the hard part was something else.
The hard part is saying no
The hard part was having someone in the room with the authority to say: we don't announce this until it's real.
That sentence sounds easy. It is one of the hardest sentences to say inside a large company. There's a keynote date. There's a competitor shipping AI features weekly. There's a product line that needs a story this year. Every incentive in the building pushes toward announcing the promise now and backfilling the product later. The only thing standing against all of that is one person with enough authority, and enough spine, to hold the line.
Apple, of all companies, used to be that person. The discipline to keep things in the lab until they were ready was the brand. It's the same muscle behind scaling a product down without breaking what made it good: restraint as a design skill. Announcing Siri twice is what it looks like when that muscle goes slack.
A demo is a promise
This is the line I'd carve over every product launch.
A demo is a promise. When you put something on a stage, you are not showing what the technology can do in ideal conditions with a hand-picked prompt. You are telling the person watching: this is coming, and it will work like this, for you.
Apple made that promise in 2024. Two years and $250 million later, it's still just a promise. The demo was real. The product was a forecast.
Sora made the same move, a demo with a download button mistaken for a product. Figma is shipping AI features faster than anyone can find a use for them. The whole industry has blurred the line between "we showed it" and "you can use it," and a demo is quietly doing more and more of the work a finished product used to do.
What it actually costs
The $250 million is the cheap part. Apple can absorb that before lunch.
The expensive part is the same thing the subscription traps and the abandoned AI launches spend without noticing: trust. Every time you announce something that doesn't arrive, you teach people to discount the next thing you announce. The promise stops landing. The keynote becomes a wish list. And for a company whose entire premium is "it just works," teaching customers to wait and see is a slow, expensive kind of damage.
Ship it, then announce it. The order is the whole discipline. A real product needs no asterisk, no "this fall," no "not in Europe." It just works, which is the thing Apple was once the only company willing to wait for.
---
