Thanks to broadly understood automation—from AI agents to the entire CI/CD process—software development has become incredibly fast. It seems that bringing the most diverse, sometimes completely crazy ideas to life is almost too easy today (This isn't entirely true, but I don't want to focus on that in this post).

I used to have a website... I wonder if I can easily revive it?

Two days later, with the help of an agent, I have a recreated and refreshed frontend, a ready mechanism for publishing a static version of the site, and automated translations. It turns out that with a properly designed harness (the entire environment built around agents, tools, skills, memory, context, and a well-defined way of executing tasks)—an idea and a little bit of steering the model are often enough. Thanks to safeguards and proper separation of privileges, I even let the agent migrate the domain to a new DNS provider. This experience shows just how much has changed since the "ChatGPT moment."

Programming feels more and more like bringing my own thoughts into reality.

This ease of creation also means that I take immense pleasure in pursuing all my whims and little "nice-to-haves" that used to cross my mind, but were never worth several days or weeks of work.

A new category is starting to emerge: One-Person Software. The cost of creation has dropped so significantly that, perhaps for the first time, it is becoming economically viable to build software whose user count is exactly 1.

Software that doesn’t have to be a product. It doesn't need a market, a business model, or a polished interface. It can be created solely to solve one very specific problem for one person.

In our case, this takes the form of small applications, microservices, and AI agents running in our own Kubernetes cluster, communicating with each other and automating various aspects of daily life. A sort of "Life as Code."

Today, it still looks like a set of technical building blocks you have to assemble yourself. Thanks to the explosion of intelligence we are witnessing, new standards for this type of software will likely emerge quite quickly. Perhaps entire platforms to host it, or maybe we will run such software directly on our phones. I suspect a significant portion of us will build (or is already building) a personal agent or an entire system of applications and automations tailored to our own lives.

Perhaps in a few years, the personal software we use won't be a product bought from a company. It will be software built specifically for us.