Software for a Market of One_

I have built more software for myself in the past few months than at any other point in my career. Much of it would make a poor product. It reflects how I arrange work and which shortcuts I trust. Another person would probably want the same capability in a different shape. I use it every day. I have no serious reason to make it useful to anyone else.

The interesting part is how quickly this library has expanded as the cost of making each tool has fallen. I once had to justify a small utility against the setup and maintenance it would create. An agent changes that calculation. A narrow tool can solve one recurring annoyance. It can change when my habits change and disappear when the habit does.

This is personal software in the literal sense. Its addressable market is one.

A workshop jig holds a piece in exactly the position needed for a repeated task. It can be immensely valuable at one bench and useless at the next. Nobody expects it to become a global product. Its value comes from fitting the work closely enough that the work becomes easier. Desktop agents are turning code into the same kind of material.

I have argued before that disposable software will occupy much the same place as spreadsheets. That argument now has another layer. A spreadsheet remains a general application into which people place their work. The agent can generate the application around the work. The useful unit shrinks from a product category to an arrangement of code around one person's habits.

This puts pressure on the personal end of SaaS. I find it harder to imagine paying for a small utility when I can ask a desktop agent to build the exact behaviour I need. The generic product carries compromises I do not need and a subscription I now have a way to avoid. Sharing the source code does not solve the mismatch. Another person may reuse the idea but their agent will probably reshape the software around their own routine.

Enterprise software follows a different economic logic. A company needs many people to work against the same state. It needs access to remain governed after the original builder leaves. The software must survive handovers and carry responsibility when something goes wrong. Code generation makes the interface cheaper but it does not remove that institutional burden. The subscription pays for continuity as much as capability.

This is why I do not expect SaaS to disappear from corporate work. The durable businesses will carry burdens that code alone cannot. They will own the layer that persists when a generated interface is replaced. For anyone building software for others, durable value is moving away from generic personal workflows and towards systems that must outlive the person using them.

I am probably early because I already know how to work with code and agents. That advantage will shrink as desktop agents get better at turning intent into a working tool. Personal productivity software will become less like a category we shop in and more like something we shape around ourselves.

Much of it will never need another user.