I want to say up front that I think it is genuinely the coolest thing in the world that some of you are building software now. The rest of this post might read like criticism otherwise — it isn't. Twenty years ago, the idea that a non-technical CEO would sit down on a Saturday morning and build a working internal system for the business would have been a joke. Even five years ago, the closest you could get was a low-code platform that mostly meant "this almost works for the simplest thing imaginable, but the moment you need anything real, call a developer." That ceiling was real, and the ceiling is gone.
I watch operators do things in Lovable, Bolt, Replit, and Claude Code now that I would have hired a $400K-a-year senior engineer to attempt a couple of years ago. Not toy versions, real ones. A construction operator with a multi-state job-cost rollup their CFO actually trusts. A boutique law firm with an intake-to-conflict-check workflow that runs in production. A practice manager who automated insurance verification end to end and freed up a full-time human's worth of work. These aren't impressive-because-an-amateur-built-them. They're impressive, full stop.
This is, in my opinion, the single biggest leverage shift for small and mid-sized businesses in twenty-five years. Not AI in the abstract — this specific thing, the ability for the person who actually understands the business to author the software the business runs on. We should all be cheering for it. I am.
And — because you knew the "and" was coming — there's a part the tool didn't tell you about.
The tool let you skip past the part that used to be hard, which was writing the code. It didn't skip past the part that's actually hard, which is operating the code. Those are different problems and always have been. Most of the cost and most of the risk of business software, historically, has lived on the operating side, not the building side. Here's the dirty secret of senior engineers — and I say this as somebody who's hired plenty of them — most of what the great ones earn their salaries doing is operations. Architecture, observability, deploy safety, incident response, the thing-that's-broken-at-2am skills. Writing code is a smaller fraction of the job than the title implies.
So when AI commoditizes writing code, the cost of building drops to almost nothing, and the cost of operating stays exactly where it was. The ratio flips. It used to be that you spent 80% on building and 20% on operating. Now you spend five hours and a Lovable subscription on building, and the entire iceberg of operating is right where it always was. Most builders don't see that until the first time the thing breaks at an inconvenient moment.
The trap for the leader-builder isn't that they fail to operate the system. It's that they succeed at it, just barely, and burn the wrong attention doing it. The cost of a CEO doing a fix on a Tuesday night isn't the fix. It's the strategic work that didn't happen on Wednesday because they're tired and re-reading error logs in their head.
I don't want fewer operators building software. I want a lot more. I want the construction founder to keep shipping rollups, I want the boutique partner to keep building intake flows, I want the practice manager to keep automating things that were previously thirty hours of human time a week. That's how this generation of small businesses gets a real productivity story to tell.
What I want is for the part that comes after — the part where the tool becomes load-bearing and starts demanding ops discipline — to not fall on the same person who built it. Different job. Different shape. Doesn't need to be the founder. Honestly shouldn't be. Build the thing. Be proud of the thing. Show your kids the thing. Then hand the operating side to someone whose actual job that is, and go run the business.
That's the deal we should be making with this generation of tools. Not "AI replaces the executive's developer." More like "AI lets the executive be the architect, and frees them from being the on-call engineer." The first half is happening. The second half is what we're trying to make real.