the terminal is a power tool, not a product

a teammate who lives in the terminal wanted one inside our app. i said the job is to give that power to people who would never open a terminal. and stickiness is our problem.

share x linkedin facebook

a small warm hill with a lone figure holding a lantern on top, a boxy robot standing beside it, and soft clouds drifting over an open sky

on wednesday we had one of those product debates that sounds small and isn't.

one of my partners does almost everything in the terminal now. ai coding agents, voice input, dozens of tasks at once. he pitched putting a terminal window inside the app we're building, so power users like him wouldn't have to leave.

i said it was doable and not a bad idea. then i asked a question back. what can the terminal do that our app can't?

we worked through it. the one real difference is access to the whole computer, and our users don't need that to run their sales and outreach. he agreed. he wanted the terminal because it's his habit, not because customers need it.

that's the trap for anyone building with ai right now. the builders live in the terminal, so they assume the terminal is the product. but for most people it's intimidating to look at and terrible at keeping context. the job of the product is to give regular users the power of the terminal in a form they can see.

the second part of the debate was more uncomfortable.

he told me the web app isn't sticky. he has fifty browser tabs open and the app gets lost among them. a desktop version helps, but it needed more work, so i asked him to use the web app every day in the meantime.

and then i told him, and the team, what i think about stickiness. if he doesn't open it, that's our problem, not his. a user who forgets your product isn't lazy. your product didn't earn the spot.

so here's the loop i asked for. keep the app open. every time you reach for another tool instead, write down what you were trying to do. that's the next thing we build or fix.

he pushed back in a useful way. even the tool our customers already use every day isn't sticky enough for him. his ideal is an agent that does one useful job right, every time. win at one thing before you try to win at everything.

he's right about that, and it's why the latest version is narrower. the first version tried to do too much. we cut toward the core, the workflow customers already run every day, then pull in the handful of things they leave the app to do. calendar. prep notes before a call. the inbox. the numbers that tell you whether the week is working.

i've written about building for habit before. this is the other side of it. you don't get habit by adding. you get it by noticing every moment someone leaves, and closing that door.

long term, i think ai products will be judged by how often people return to them without a reminder. clever features are cheap now. the daily return is the scarce thing.

more on how i build with ai is in the ai book.

so the last time someone on your team skipped your product for another tool... did you ask them why?

share x linkedin facebook
the newsletter

the machine economy brief

one email when it matters: bitcoin, ai, robotics, and what founders should do about it. unsubscribe anytime.

no spam, no list-selling. your email goes to sunny, nowhere else.

this is the thesis i'm building on.

i write the machine economy brief: bitcoin x ai x robotics, where it's all going, once a week.

join the machine economy brief →