Something changed in how software gets built

For most of software history, the thing that slowed projects down was how much code people could write. So you hired more people.

AI coding agents flipped that. Writing a component, an endpoint, a migration, or a bunch of tests now takes minutes. What's scarce now isn't code, it's judgment: knowing what to build, how it all fits together, and whether it actually works.

More people, more back-and-forth

Here's the thing nobody likes to say out loud. When you put five people on one product, a big chunk of the week goes into keeping everyone in sync. Meetings, tickets, handovers, waiting on someone else, and redoing work because two people solved the same thing differently.

When code was slow to write, that was worth it. Now that AI agents write a lot of it really fast, the syncing becomes the slow part. Adding people can actually make a project slower, not faster.

  • Requirements get passed along and lose detail every time.
  • Frontend, backend, AI, and data get designed separately and only meet at the end.
  • Bugs hide in the gaps between people, where nobody feels responsible.

One person holding the whole picture

When I build something, the interface, the API, the database, the LLM workflow, and the data integrations all live in one head: mine. They're designed together, so they fit together.

That's honestly where most of the quality comes from. Not from more reviewers, but from one consistent way of doing things across every layer, and nobody guessing what the other side expects.

What it costs you

A full-time team costs a lot more than the code it writes. Salaries and benefits, recruiting, onboarding, managers to coordinate everyone, meetings, and the weeks where there just isn't much to do but you're still paying.

And here's the part people don't expect: it's not just cheaper than a team. Working with me as a contractor is often cheaper than hiring even a single full-time developer. You mostly pay for the actual building, and instead of someone who covers one part of the stack, you get someone who covers all of it.

  • No recruiting or onboarding. We can start pretty much right away.
  • No management layer. You talk directly to me.
  • No paying for quiet weeks. You pay for the work.
  • One person for the whole stack, not just one part of it.

How I actually work day to day

I use AI coding agents for the volume: scaffolding, boilerplate, tests, refactors, and first drafts of features. My time goes into the parts that need experience: architecture, data models, security, checking the AI's output, and reviewing every change before it ships.

In practice that means you can tell me about a problem in the morning and often see something working the same day, built by the same person you talked to.

When you really do need a team

I'll be honest: if you're running a huge platform, need people on call 24/7, or have lots of separate products going at once, you need a team. I'm not a replacement for an entire engineering department.

But for most companies, one experienced person with AI tools can now handle all of their software: new AI products, AI in what you already have, internal tools, and keeping it all running. It's often faster, cheaper, and better than hiring. That's exactly what I love doing, so if that sounds like you, send me a message.