Last Thursday OpenAI finally shipped an actual iOS app for ChatGPT, and I downloaded it approximately four minutes after I saw the tweet, standing in my kitchen at like 9:40pm waiting for a kettle to boil. Free download, no ads (for now, and I do mean for now), syncs your chat history across devices, and if you're already paying the $20 a month for Plus it hands you GPT-4 instead of the slower 3.5 model. It's US-only at launch. Android users get told, again, to wait, which at this point is basically Android's whole personality when it comes to hot new apps. I don't say that to be smug, I had an Android phone for three years and got that same "coming soon" treatment more times than I can count.
The thing that actually got me, though, wasn't the app itself, it's basically just a wrapper around the website with a nicer keyboard experience. It's the voice input. They built it on Whisper, their own speech recognition model, and it is genuinely the best dictation I've used on a phone, better than Apple's native mic button by a wide margin, especially with any kind of background noise. I tested it against my dishwasher running and it didn't miss a word. My dishwasher is loud. It's from 2014 and sounds like it's grinding rocks.
So now I've got this app on my home screen and I've caught myself, three separate times in five days, talking to it out loud like it's a person standing next to me. Not typing. Talking. "Hey can you rewrite this email so it doesn't sound annoyed" — spoken into my phone in my car, in a Trader Joe's parking lot, with the windows up so nobody would think I'd totally lost it. It rewrote the email fine. I still sent the annoyed version because sometimes you want the email to sound a little annoyed, that's information too, but that's beside the point.
I want to be clear that I'm not writing one of those posts about how this changes everything forever. Plenty of people are already writing that post this week and I don't think I have anything to add to it that's actually true. What I noticed instead is a smaller thing: the interface changes how you use the tool. On the website I'd type out careful, edited prompts because typing on a laptop invites editing. On the phone, talking into it, my prompts are messier, more conversational, more like how I'd actually phrase a question to a coworker. And weirdly the answers feel more useful that way. Less like querying a database, more like asking a person who happens to know a lot and never gets tired of you.
There's a context thing worth mentioning too. Sam Altman was in front of the Senate a few days before this app dropped, telling Congress that AI probably needs some kind of licensing regime, that his own company's technology could "cause significant harm to the world" if it goes wrong. I watched maybe twenty minutes of the hearing clips on my lunch break. It's a strange thing, watching a CEO ask lawmakers to regulate his own product more, and then five days later shipping that same product straight onto three hundred million iPhones with zero friction. I'm not accusing anyone of hypocrisy exactly, I think both things can be sincere at once, a company racing to ship while genuinely hoping someone builds guardrails before it's too late. But it's a weird two-step to watch happen in the same week and I don't think enough people are pointing at how weird it is.
Anyway. Practical notes for anyone downloading this today: it's a big download for what it is, the app crashed once for me on launch day and hasn't since, and if you've got a Plus subscription already tied to your account it just recognizes you, no extra sign-in nonsense. The chat history syncing back to the website is the actual killer feature for me, I've been copy-pasting prompts between my phone and laptop for months like some kind of caveman and now I don't have to.
I still think the dishwasher needs replacing more than I need a chatbot that talks back to me in a Trader Joe's parking lot. But here we are.