Company of One

A two-minute deploy that takes two hours isn't a deployment problem. It's a maintenance problem.

The server hasn't been touched in months; the dependency chain is out of date, the distribution updates haven't run, and by the time you've untangled it all, you've lost the evening. Every change needs to run on every server — and there's always something you forget. They're never quite in sync. So whenever you need to make a change, you're not just deploying; you're catching up first.

It's also never the right moment. Server maintenance doesn't get planned. It finds you.

I was running this across my own servers and client servers at the time. Ubuntu 16.04 running PHP 7.0, no way to run a current Laravel without a full OS upgrade — which breaks something, which needs debugging, which costs the time you didn't have. Pretty standard. Pretty much every server I've ever touched.

Building a tool for yourself is easy. Building a product for others means you have to actually finish it.

strackt started as an itch. I was tired of the maintenance loop — so I started building something that fixed it for me. That was enough for a while. But at some point the question shifted: do I keep this as a personal tool, or do I build it into something real?

Turning it into a product meant different standards. It couldn't just fix my problem; it had to fix it well, reliably, for someone who wasn't me — at a moment when I wasn't there to help. I wasn't sure I had the resources to pull that off alone.

Reading Company of One by Paul Jarvis didn't give me a plan.

Company of One isn't really a business book — it's a reframe. Paul Jarvis's argument is that staying small isn't a failure to grow; it's a deliberate strategy. That growth for its own sake is a trap. That a business built carefully, without investors or headcount, can be more resilient than one that scales before it's ready.

It covers a lot: how you run a company, how it fits into your life, what you actually need versus what you assume you need. I read all of it with strackt in my head — not looking for a blueprint, more looking for reasons it could work. It gave me enough. I decided to at least try building it properly. Didn't know if I'd finish it. But if I did, it would need to be able to run without me.

I took two things from it: focus on the user; automate me — both ended up baked into strackt at the foundation.

"Focus on the user" means the product isn't built from my perspective — it's built from yours. No technical jargon. No configuration screens that require you to already know what you're configuring. Simple wizards, progressive disclosure, the server doing what it's supposed to do without you needing to understand how. You shouldn't have to think about what's happening underneath; you should just see that it works.

"Automate me" is how I think about my own work in general: how do I make myself redundant? For strackt, that means automatic updates, automatic patching, automated management. The product handles the maintenance loop so that nobody — not you, and not me — ever has to. If you run into a problem and solve it without ever contacting support, that's the win.

The book convinced me it was possible. It didn't tell me how to be good at things I'm not good at.

I'm a developer. I know what good looks like — a clean interface, a well-placed tooltip, a workflow that just makes sense. I know what doesn't work. What I don't always know is how to make it. Design, visuals, the creativity side of building a product — not my strengths. And as a company of one, there's nobody to hand that off to.

I had the conviction. I had the technical foundation. I sat with that gap for a while.

The company-of-one model always made sense. What changed is that the tools caught up.

AI handles the parts I'm not good at — design direction, visuals, the creative gaps — well enough that I can move forward without a team. Not perfectly; there's always a moment where you can tell it didn't quite get it. But without it, there's no product.

It's the same logic as strackt. You like developing; you don't like managing servers. strackt covers that gap. I like building; I just can't design. AI covers mine.

The conviction was always there. Now so are the tools.