The Age of Personal Software_

Some of the most useful software I use now has a market of one. It is me.

Until recently, that would have been a difficult way to justify building it. Any software product, serious or not, needed enough time, money, and maintenance that it had to serve a broad group of possibly paying customers. The economics pulled every idea towards a SaaS business because of this.

This constraint produced valuable companies. Needless to say, it may have also filtered out an entire category of useful software for tools that are too specific to become businesses, but useful enough to make a person's day better.

Over the past 7 months, AI agents have been lowering that threshold. They can help turn a small, well-formed need into working software while the need is still fresh.

This does not make engineering free, and it does not make judgement optional because small tools can still be wrong, insecure, or annoying. What changes is the amount of justification they need which is practically none at the moment.

Scry is a good example from my own work. I wanted a simple way for an AI agent to turn the result of an investigation into something I could inspect visually. The agent can prepare and validate a small result set, write it as CSV, and pipe it into Scry. A native chart appears in its own macOS window. I look at it, close it, and continue with my work. Works perfectly with voice mode in ChatGPT.

Scry is a single Swift source file built on frameworks already present in macOS. It came together quickly because it does not try to become a general analytics platform. It handles one narrow handoff well.

That narrowness is the point. If I had started by asking how to monetise Scry, I would have designed a different thing. I would need to think about teams, storage, permissions, integrations, onboarding, and every edge case required by an unknown customer. Those may be sensible requirements for a commercial product, but they would have buried the small utility I actually needed at the time when I needed it.

Here is the workflow I find interesting and it is quite simple. Notice friction in daily work, ask an agent to help build the missing piece, and lean on native capabilities wherever possible because it has the lowest friction. Then use the tool in the real situation that created it. If it improves the work, keep it. If it does not, change it or let it go.

Software has always included personal scripts, macros, and small automations. Every engineer had made their own dotfiles at some point right? What feels different now is that this is coming for the masses. A personal tool can have a proper interface, behave like part of the operating system, and be pleasant enough to become a habit without first becoming a startup.

I am not arguing against SaaS or commercial software. Some problems genuinely need shared infrastructure, collaboration, guarantees, and a company standing behind them. Personal software expands the other end of the spectrum. It makes more ideas worth trying because usefulness to one person can be sufficient.

Sometimes one of these tools will turn out to be useful to other people. That is a welcome bonus but would this first make my work better?

That is why I think the age of personal software is here and not in a form of SaaS. We can build for a market of one, at the moment a need appears, using capabilities already sitting on our machines.