You used to be able to think.

It happens slowly.

You write fewer emails on your own. You ask before you decide. You summarize instead of read. You navigate by GPS to places you’ve been to twenty times. None of it feels like a problem until you sit down to do something you used to do without thinking about it, and the cursor just blinks.

A friend told me last month that he hadn’t written an email himself in three weeks. He didn’t mean a polished draft. He meant the words, the order, the decision about what to say. After three weeks of letting a model take a first pass at everything, he sat down to write a short message to his sister and didn’t quite know how. He laughed when he told me. He also said the laugh was the bad kind.

This is happening at scale, and it is mostly unobserved by the people inside it. People who have read books their whole lives can’t remember what they read yesterday because they summarized it. People who used to make hard calls now consult three different models first. People who wrote love letters wonder if anything they say sounds like them anymore. The capacity isn’t gone. It just hasn’t been used.

Quitting.ai exists for the people who notice this.

The premise is small. Certain cognitive capacities are use-or-lose: your writing voice, your judgment, your memory, your sustained attention, your originality, your social bandwidth, your everyday practical sense. They got built up over a lifetime of small, unaided efforts: writing the email yourself, doing the math in your head, walking to the meeting without checking the route, sitting alone with a problem until something clicked. Outsourcing those efforts to a model doesn’t destroy the capacity. But it does stop you from exercising it. After enough time, the use-or-lose ratchet does what it does.

We are not anti-AI. We use it. We will use it tomorrow. The tool has real and durable uses, and pretending otherwise would be silly. But the early evidence is that a lot of people are using it in ways that are quietly eroding things they cared about, without quite noticing. The point isn’t to quit. The point is to use it on purpose: to know what you’re trading, and to keep the capacities that matter to you in working order.

This is also why the product has an AI in it. The Report’s reflection is generated by Claude. We thought about whether to do this for a long time, and decided that the alternative, a generic template, would treat the reader less seriously than the brand promises to. So we use it once, on the read of an Audit, in a place where personalization adds something a template can’t. We don’t use it for the Audit itself, the daily tasks, the streak math, or anywhere else that doesn’t earn it. That’s the line we drew. It might be the wrong line; we’ll see.

What we’re trying to build is not an app you spend time inside.

It’s a quiet program that lives mostly in your inbox. One short task each morning, asking you to use a capacity you’ve been outsourcing. One longer task on Sunday. A two-minute reflection on Friday. A re-Audit at thirty days that shows you what changed. Then we’d like you to leave and live, and check in next month.

A capacity, once lost, can be relearned. Most of the work is showing up, without help, for a small thing each morning, and noticing what comes back.

You used to be able to think.

Let’s get that back.