OpenAI's latest price cuts to its API landed with the dull thud of a corporate spreadsheet update.
No keynote, no celebrity demo, just a blog post and a new number per million tokens.
And yet that number is reshaping the apps on your phone faster than any product launch this year.
For the uninitiated, the API is the pay-by-the-drink pipe that lets developers plug OpenAI's models into their own software.
When the price per token drops, the cost of every AI feature built on top of it drops with it.
Translation, summarization, customer chat, photo cleanup, voice notes, homework help — all of it gets cheaper to run overnight.
That sounds like good news, and in the narrow sense, it is.
But it also means the floodgates open wider.
Features that were too expensive to ship last year suddenly pencil out, which is why your banking app, your grocery app, and your kids' school portal are all about to sprout an AI assistant whether anyone asked for one.
Here's the part nobody puts in the press release.
Cheap intelligence doesn't stay confined to helpful tasks.
It gets bolted onto everything, including the things that worked fine without it.
Your email client now offers to rewrite a two-line reply to your neighbor about a fence.
The economics push in one direction only.
When a feature costs a fraction of a cent per use, companies stop asking whether users want it and start asking why they haven't shipped it yet.
The default flips from opt-in to opt-out, and the burden of saying no lands on you, one settings menu at a time.
Small developers and solo builders can now afford to experiment with tools that used to require venture funding.
A high schooler with a laptop can ship a study app that would have cost thousands a month to run two years ago.
That kind of access is genuinely new, and it's worth defending.
The same low prices make it trivially cheap to generate spam, fake reviews, and convincing scam messages at industrial scale.
They need a credit card with twenty dollars on it.
Every price cut is also a subsidy for the worst actors in the room.
And then there's the quiet erosion of craft.
When a machine can draft, summarize, and respond for pennies, the pressure to slow down and do it yourself evaporates.
We get more output and less thought, more replies and fewer conversations.
The cost savings are real; so is what we trade away for them.
None of this is a reason to torch the data centers.
It's a reason to pay attention to what gets built now that the meter runs cheaper, and to notice when a feature exists because it helps you versus because it was finally affordable to bolt on.
The price of AI is falling, and that's a genuine shift.
Whether it makes our tools better or just noisier depends on choices companies are making right now, mostly without asking us.
Final Thoughts
Cheaper isn't the same as wiser, and we should stop pretending otherwise.