Personal software has become cheap to build. With a coding agent and a few prompts, I can make a small tool that does exactly what I need, like a subscription tracker that reads my bank export and flags what I forgot to cancel, a habit app, in an afternoon. And it isn't only toy apps anymore. A personal agent can fork an open-source project and bend it to my needs, wire up a sophisticated library, or stand in for a developer tool I used to pay for. So I've almost stopped paying for software subscriptions, small or not.

Most consumers can't do this yet. That isn't because the building is hard anymore. It's because the setup is: installing Claude Code, writing a few prompts, then deploying the result somewhere you can actually use it. That's still beyond most people, so they keep paying a monthly fee for an app that tracks their subscriptions, which is a subscription to manage their subscriptions.

I think that gap is about to close. The next generation of devices will have agents built in, living close to the user, with access to their context and the ability to build, run and host software for them. When building an app is as easy as asking for one, a lot of the app store's value goes away.

A few things follow from that:

1. Building has no friction. You describe what you want and your agent builds it, runs it and keeps it updated.

2. Sharing beats selling. People will share the tools they make the way they share a recipe, not list them on a store.

3. Anyone can fork. If someone's tool is 90% of what you need, your agent adapts the last 10% for you. Forking stops being something only developers do.

Heavy software, like games, anything built on network effects, and anything trusted with money or health, will be the most robust category and the last to go. But even there, I can see the model changing. Instead of a company building one product and selling it to millions of individuals, a group of people with shared or complementary needs could have their agents coordinate to build and maintain it together, as one general solution that serves the whole group. The long tail of small single-purpose apps, which makes up a huge share of what people pay subscriptions for, will go first. The heavy apps may just take longer to follow.