For most of my career, when a friend or family member came to me with a software idea, my answer was the kind of soft "yeah, maybe someday" that everyone knew really meant no. A real custom build is a hundred hours of unglamorous work (auth, hosting, deploys, edge cases, the boring 80 percent), and a hundred free hours wasn't something I had to give. So the asks piled up. I wanted to help every time, but the math wasn't there, and nothing actually got built.

That started to change in early February, when the latest round of models hit. Two pretty different things shifted at once.

I needed somewhere to actually use this stuff

The pace of AI right now is the fastest I've ever watched a technology move, and I've been doing this long enough to have seen a few inflection points. New models, new tools, new agentic patterns. The ground under what's possible keeps shifting week to week. You can read about all of it and still not really know how it works. The only way I've ever learned a new technology is by actually building something with it.

The catch is that my day job, like a lot of jobs in regulated or risk-averse industries, has real limits on where and how AI can be used. There are good reasons for those limits and I'm not complaining about them. But it does mean the slice of AI I get to touch professionally is small compared to what's actually out there. If the rest of my exposure comes from reading what other people are writing about it, I'll always be a year behind whoever's actually shipping.

So I needed somewhere of my own. A place where I could take an idea all the way through: pick the model, wire it up, deploy it, watch it break, fix it, do it again. Side projects do that better than anything else I've found. A weekend on something real teaches you more than a stack of conference talks.

And the math finally adds up

For my whole career, the productivity ceiling on one engineer working alone was basically fixed. You got a little faster with experience, a little faster with better tooling, a little faster with a focused weekend, but the curve was gentle. The hours a real custom build needed cost more than I had to give.

AI tools have started to break that ceiling, and not by a small amount. GitHub's research on Copilot a couple of years ago put it at about 55% faster, and the current generation of models is in a different category. If you've read studies suggesting AI tools don't help much, or even slow developers down, you're reading studies that measured last year's tools. The jump that happened in early 2026 is real, and on the kind of work I'm actually doing on Saturdays it's much bigger than even those older numbers suggest.

New projects from scratch, full-stack, solo, with my tooling dialed in and no legacy code in the way. That's where these tools are at their best. I'm a heavy user of them and I've put a lot of unpaid hours into learning to drive them. The multiplier I'm seeing is 3-4x, sometimes more on a project that fits the tools well. A weekend now does what a two-week sprint used to.

AI absolutely makes you faster. But it's a multiplier, not a substitute, and it amplifies whatever you bring to it. A 3-4x boost on top of twenty years of production engineering experience looks like real software shipped in a weekend. The same boost on a developer who doesn't already know what to build, how to structure it, or how to keep the model on the rails just produces bad code faster. The tools are remarkable. They still need someone in the chair who knows what they're looking at.

So the project that used to take me a hundred hours takes about twenty-five. A friend asking for a small custom tool isn't a six-month favor I have to politely decline anymore. It's a real deliverable on a real timeline at a price that makes sense for both of us.

Why a company, and not just a hobby

I thought for a while about just keeping this as favors and never putting a name on any of it. Two things eventually changed my mind.

One was that the asks kept coming. Friends tell their friends, and "I know a guy" travels fast in small business circles. Without a real name and process behind the work, every project ends up being its own one-off scramble. With one, I can take on more without losing track of things in my inbox.

The other was that writing the code itself is rarely the hard part. What people actually need is code that's scoped fairly, hosted somewhere it isn't going to disappear, and maintained by someone they can actually call. That last part is rarer than people expect. Plenty of freelancers will build you a thing and walk away. The site goes live, the contract closes, and six months later when the database fills up you're calling someone new who has to start from scratch.

Wallman Solutions is built to be the opposite of that. Every project I take on is built, hosted, and maintained by me, the same person who scoped it. AI is what makes it possible for one engineer to cover that much ground. But the reason it's worth doing is that most small businesses don't actually need a small team. They need one competent person who picks up the phone.

If you're reading this

If you're someone I've told "yeah, eventually" to in the last few years, the answer's different now. Send me what you were thinking about. There's a good chance it's something I can actually build.

And if you're running a small business with a process held together by spreadsheets, or you've got an AI-generated prototype that looked great in a demo and is starting to fall apart in production, that's the same kind of work and I'd love to hear about it.

Drop me a line.

— Nathan