Hi. I'm Andrew.
If you're here, you probably clicked a link, and the link probably didn't say much about who's typing this. So before anything else: a short version of who I am, what I've spent twenty-five years doing, and what I want this blog to be.
The short version
I've been writing software professionally for twenty-five years. That's long enough to have built things in almost every shape of company I'm aware of — global financial institutions where a deploy was an act of paperwork; tiny startups where a deploy was a git push and a held breath; pretty much everything in between. Some of those years I was head-down on a keyboard. Some I was on call at 3am, alone. The last few I've been leading a team of engineers, all of us moving — together — into the world of building with AI.
The throughline, if I had to pick one, is architecture. Not the buzzword — the actual thing. How a system holds itself together, where the seams are, which decisions you can change later and which ones you can't. I'm interested in that across scales: a microservices fleet, a three-person team, a single function someone else has to read in six months.
Along the way I co-founded a healthcare AI startup before "agentic" was a word anyone outside research used. We were building agents that talked to insurance companies on behalf of doctors. The company didn't survive — most don't — but it taught me, expensively and clearly, what coordination between specialized experts actually costs. I think about that lesson almost every day now, because it translates with eerie precision to the work I do with AI agents.
Why I'm writing
For most of the last few years I didn't write anything. The space was moving too fast, everyone else was writing about AI, and most of it was either breathless hype or breathless doom, with very little in the middle that sounded like an actual builder talking.
But the space has settled enough now — not slowed, settled — that I have things to say I don't see other people saying:
- The tools have gotten genuinely, embarrassingly good. Most developers I know are still using them like fancy autocomplete.
- The leap from "AI helps me code" to "AI does work on my behalf" is bigger than the leap from no-AI to AI was. And almost nobody is talking about how to make that leap well.
- The right mental model for working with agentic systems is closer to coaching a team than writing a program. That has consequences, and the consequences are interesting.
- Twenty-five years in, a lot of the lessons that mattered most in 2001 still matter in 2026. Some of them matter more. Some of them are now wrong in ways that surprised me.
Those are the kinds of things I want to write about.
What this blog is
A personal tech blog. Specifically:
-
AI, agents, and how to actually build with them. This is where I'm spending most of my time and effort right now, and where I've most strongly come to believe that real working knowledge isn't widely shared yet. Agentic systems, prompting, tool choice, the design patterns that are emerging, the failure modes I've hit, the wins I've had.
-
Building software, twenty-five years in. Architecture decisions and how I think about them now versus how I thought about them in 2005. The seams between specialties. Senior-engineer instincts that turn out to apply to AI work, and ones that don't.
-
Tools I actually use, honestly. Not a curated list. The real one — with the wins, the times they bit me, and the workflow shifts that compound.
-
Occasionally, the parts of my life that have shaped how I think. I was a competitive strongman for a chunk of my thirties — multiple top-5 finishes at America's Strongest Man, a few international meets. I was diagnosed with an aggressive leukemia in 2016 and spent six and a half years in treatment, including two stem cell transplants. I co-founded a startup, made hard calls, and eventually sold it for what it was worth. I have a toddler and a wife who both impress me daily. None of that is the subject of this blog, but all of it shows up in how I think about engineering, and I'll write about the intersection when there's something real to say.
What it isn't
It isn't a product funnel. There are things I'm building — there always are — and they'll show up here when they're ready. They won't be, or at least rarely will be, the reason any individual post exists.
It isn't a "personal brand" play. I'm a person; I have opinions; that's the brand.
It isn't going to be daily. I'd rather write one thing on Tuesday that's worth the read than three things a week that aren't. The rough plan, for now: a long-form piece on Tuesdays, a short observation on Thursdays, and maybe a drop week once a month so the bigger pieces have room to breathe — and I still have time to throw some iron around.
A note on where this lives
Everything posts here first — at codelifter.net. I'll syndicate to LinkedIn and a couple of other places after the fact, but if you want to read what I'm actually writing (including the rougher versions before they get smoothed over for a different audience) this is the canonical home.
If you want each post in your inbox, the signup is at the bottom of the blog page. It's a list I own, sent from an address I control, with a one-click unsubscribe that actually works. No tracking pixels, no behavior emails, no "we noticed you opened the last one" follow-ups. Just the posts.
I have feelings about most of the alternatives. We'll get there.
See you Thursday
The next post is short. I'm going to write about something I keep saying out loud to other engineers, which is that the single most underrated skill in 2026 is writing a prompt you'd let someone else read.
After that, things get more interesting.
Let's do this!