Building personal software is now Practical

When building is cheap enough, more problems are worth solving

For a long time, there has been a gap between "I have a problem" and "I can justify building software for this". If the problem was too niche, the answer was usually to live with a bloated professional tool, pay a subscription for features you barely use, or keep doing the task manually.

Closing that gap is, I think, the biggest impact agentic coding will have on software. AI-assisted development doesn't remove the need for engineering judgement or product taste, but it does collapse the time it takes to get from an idea to a usable tool. More problems now fall into the range where it makes sense to just build something.

I don't just mean one-off scripts. I mean real applications, shaped around a specific workflow, built and maintained by the person who uses them.

A concrete example: Photograph

Over the last two days I built Photograph, a native Rust desktop photo browser and editor for my own photography workflow.

I'm a hobbyist photographer. I don't need every feature in a professional photo suite. I do need a workflow that fits the photos I actually take: browse folders quickly, inspect RAW files and metadata, make non-destructive edits, crop and straighten, apply keystone correction, adjust exposure and white balance, work with selective colour, and export with resize and compression controls.

Building this without AI assistance would have taken me months of time that simply didnt balance with the necessity. With it, the project went from first commit to published on GitHub in a weekend. That included RAW decoding for five camera formats, an interactive crop tool, colour grading, GPU-accelerated previews with CPU fallback, and packaging for both Linux and macOS.

Why build an editor?

Selective colour was the specific thing keeping me on a paid licence. If you've ever complained about photo editing on Linux, someone has inevitably told you that GIMP and darktable are perfectly good alternatives and that people who pay for Lightroom are just being wasteful, but while Open source photo editing tools are almost always technically impressive under the hood, the features that matter most to me in my editing workflow (selective colour, keystone correction) either don't exist or are buried well enough that they might as well not. Compare that to Lightroom or Capture One, where these are first-class controls, and you can see why people pay for subscriptions. They're the kind of features you only know to build or surface if you've sat there trying to pull a specific hue out of a sunset.

Photograph has selective colour because I needed it, and it's central to the workflow. It's not a replacement for Lightroom or Capture One, but it's software shaped around how I actually edit photos, and I don't pay a subscription for it.

What actually changed

The shift isn't just that code gets written faster.

The shift is that exploration is cheaper. Trying a UI idea, refactoring a pipeline, adding a missing feature, testing an alternative architecture, pushing through the tedious parts. All of these got cheaper at once.

When exploration is cheap, more niche software gets built. That matters because most of our real workflows are niche. A photographer's workflow, a writer's tools, a researcher's pipeline, a one-person business. We've spent years forcing these into general-purpose products because custom software was too expensive to justify for a small audience.

That trade-off is now ready to be re-evaluated.

Why I wouldn't have built this before

One thing I've noticed over the years is that many types of software require a surprising amount of domain knowledge just to get something basic working. A photo editor is a good example. Even if you only want to crop and adjust exposure, you still need to deal with RAW decoding across multiple vendor-specific formats (RAF, DNG, NEF, CR2, ARW, each with their own quirks), colour space conversions, non-destructive edit pipelines, and GPU acceleration to keep the interface usably responsive. The knowledge required is not weeks of reading. It is years of specialisation in areas most people will never need for anything else.

We cannot know everything. At some point you have to be deliberate about what you spend the finite hours of your life learning deeply, and "GPU rendering pipelines so I can browse my holiday photos" does not make the cut. Before, that's where the idea would die. The rational move was to pay for Lightroom or Capture One, or live with something close enough.

What changed isn't that this specific deep knowledge stopped mattering. It's that I no longer have to personally carry all of it before I can start. That knowledge still exists, accumulated by the people who built the libraries, wrote the specs, and debugged the edge cases over decades. Agentic coding tools let you build on those shoulders without first having to climb up to them. I can focus on the parts where my knowledge actually matters: what my workflow should feel like, what to cut, and when something is good enough. The cost of each iteration dropped enough that a weekend could cover ground that previously took months, and that moved Photograph from "nice idea, not worth it" to "I'll just build it."

I think a lot of software is stuck in that gap. Not because people can't imagine what they want, but because the domain knowledge barrier is too wide for the audience size. When that barrier gets thinner, the software gets built.

The middle ground

This isn't a purity argument against professional software. If you're a working professional, paying for a polished tool is often the right call. But tools built for a wide audience have to make compromises, and those compromises tend to fall on the niche workflows.

The point is that there is now a practical middle ground between "toy script" and "enterprise product". Software that is too niche for a company to build, but useful enough to be worth a weekend of focused work. Software you own, that fits your problem, and that you can change when your needs change.

I expect more of this kind of thing. Some will stay personal. Some will get shared once it turns out other people have the same niche. Some will quietly replace subscriptions.

Photograph is just one example, but it already demonstrates the basic idea: the economics of building for yourself have changed, and a lot of problems that were not worth solving with software now are.

PS: If you do need professional-grade editing software, Capture One is a genuinely excellent alternative to Lightroom and still offers a one-time perpetual licence as an alternative to their subscription. We should be supporting that model wherever it exists, before the option disappears entirely.