We Are All in the C-Suite Now

A few months ago I wrote about the greenhouse and the lens, the idea that agentic AI work splits into two modes: a greenhouse where you let everything grow and pick out what’s worth keeping later, and a lens where you focus on a known target. I’ve spent every productive working day since in one of these two modes, and I’ve been thinking a lot about that essay since then. I was completely right, and I wish I had been a bit wrong.

I’d love to wave my hand vaguely at an AI and tell it to go build something amazing. But AI still can’t zero shot a startup for us, let alone a hockey stick growth company. Point a model at nothing in particular and you get a greenhouse full of plausible weeds. Someone still has to know what the target is, and that someone is a pro like me (or you).

What it can do is still remarkable. AI models now do essentially all of my hands-on-keyboard coding. For my whole career, the core of the job was sitting down and typing code. Now I almost never do. Opus 4.5 shipped on November 24, 2025. Within a few weeks of that release, I had basically stopped writing code.

I find this kind of depressing. Time in flow, writing code, meant something to me. It was fun. You know the kind of afternoon: you load a whole system into your head, the hours disappear, and at the end something works that didn’t before. Those afternoons are gone, and I don’t think they’re coming back.

What I got in exchange sounds unhinged when I say it: I can speak new products into existence. I describe what I want, sometimes out loud, and an hour later the real thing is running, with tests.

But the work still eats all my focus, just a different kind than any software developer is used to. The minutiae are gone. I don’t think about variable names, or whether a loop should be a map, or which side of an off-by-one I’m on. The agent handles that, mostly well. What’s left is the 30,000 foot view: what are we building, for whom, why this and not that, and is this still what I set out to make, or has it drifted? Those are the greenhouse and lens questions. They used to come up now and then. Now they’re the whole job.

We are all in the C-suite now. If you’re working seriously with these tools, you’re doing a CEO’s or CTO’s job: setting direction, deciding where attention goes, judging work you didn’t do, and deciding what not to do. The agent is a very capable staff that never sleeps. It will build exactly what you asked for, including the wrong thing, with total confidence. If you’ve ever watched a leadership team chase a bad strategy at full speed, you know how that goes.

Our computer science and software engineering educations didn’t prepare us for this. We’re trained to go deep and sweat the details, because that’s where the bugs lived. The career rewarded people who could hold a gnarly piece of code in their head and get it exactly right. Those of us who loved that work now have to unlearn the urge to dive in and stay up at altitude, knowing where the seams are, instead. Nothing is concrete up there, and feedback is slow. Your own judgment about what matters is the only thing between you and a pile of well-built irrelevance.

Executives have a reputation for losing touch with the work, and sometimes deservedly. We run the same risk. Stop touching the code and it gets easy to stop understanding it. Once that happens, you lose the oracle that tells a right answer from a plausible wrong one. Staying at 30,000 feet without losing sight of the ground is the skill I’m working on now. I haven’t mastered it yet, and I still miss the old one.

We’re all trying to figure out how to navigate this new reality. I don’t think anyone has the answers yet, but wherever we land, this craft will look quite different from where it was before Opus 4.5.