Use cases

Every app you've ever used has guardrails.

It does what its makers imagined you'd need — built for a million customers, none of them exactly you. Even the "AI-powered" tools you're sold are someone else's idea of what AI should do, with a fence around it.

When AI writes your software, the fence is gone.

One tool replaces all of them, because there are really only three moves — and each builds on the last:

ConnectYour data, wherever it lives BuildThe exact tools you need on top of it AutomateSo it all runs without you

Connect

Connect your data, wherever it lives

Everything starts here, because everything runs on your data — every tool you build, every chore you automate. The catch: your data lives everywhere. Some in email, some in your billing system, some in spreadsheets, some locked inside apps. This is what writing code unlocks: the agent writes you a program that goes and gets it all, pulling your information out of wherever it lives into one place that's yours — and you can run it again whenever.

Years of receipts buried in email — dug out into one organized spreadsheet
Files scattered across drives and folders — gathered into one place
Reviews sitting on five different sites — pulled into a single sheet, with the trend

Build

Make your own mini-apps

With your data in reach, build on top of it. Think about the apps you pay for or put up with — each one is somebody else's guess at what you need, close but never quite right. Now you can describe the exact thing you want, and have it. People with zero coding background are building little apps that fit their life perfectly, instead of signing up for another subscription.

A spending dashboard built straight from your bank statements
One screen with the five numbers you actually steer by, always current
A search box that answers questions from all your files — wherever they live

Automate

Hand over the busywork

The last move ties it together: anything you've connected or built can be set to run on its own. Every morning, every Monday at 8, or the moment a certain kind of email arrives — you say the schedule in plain English, and from then on the work just happens. This is where the hours come back: even Anthropic's own legal and finance teams use these tools this way.

A folder that files whatever you drop into it — read, renamed, logged
A morning brief that's ready before you are
A daily check that runs so you never have to remember it

All three together

Real-World Examples

Watch the three moves work together on a single task. Pick an example — and remember, they're just examples: this exact process works for anything you can describe, at work or anywhere else.

1 Connect

Think of the report you put together every month

The numbers come from a system export, a spreadsheet a coworker maintains, and a couple of email attachments — and you're the courier who collects it all. First move: have the agent write a small program that does the gathering — every piece pulled into one tidy spreadsheet, organized the same way every time. That program is yours now. Run it whenever; you never copy-paste it together again.

2 Build

Now put a view on top of those numbers

The data's in one place — now build on top of it. A custom report, more detailed and more tailored to you than anything a data analyst would have handed you: totals by month, this month against last, the comparison your boss always asks about, anything unusual flagged. You stopped assembling information and started reading it.

3 Automate

Now make report day disappear

Last move: put it on a schedule. The first Monday of every month, the new numbers get gathered, the dashboard refreshes, and a drafted report — in your usual format — is sitting in your folder before you sit down. The chore that used to eat half a day now happens while you sleep. You read it over, fix a sentence, and hit send.