Yesterday’s post had a section called what is Claude actually doing here, with a tidy little table splitting the labour between me and the model. A friend read the post and pushed back on it.

“At first read, it feels like you’re trying to devalue the work you do through Claude and make it seem like Claude is the smart one and you just have ‘ideas’. I would’ve preferred the opposite perspective. Claude, a tool, brings out in you and to life ideas you have that are constrained by time, not ability.”

He’s right. The table makes Claude the intelligent one and me the person pointing at things, and that isn’t true.

What stings is that I’d already written the better version. In Agentic Coding Is the Consultant’s Revolution I argued that the people who gain the most from these tools are the ones who saw problems and couldn’t build solutions, the consultants and operators and analysts and church tech teams whose ideas died in the gap between “I can see how this should work” and “I can afford the weeks it would take to build it”. The ideas were always there. The construction cost was the blocker.

Yesterday I let that framing slip into something weaker, so here’s the version I should have written.

I have ideas all the time. So does everyone. Most of mine used to die in the gap, the same way yours probably do. Tour of the Bible lived in my head for months before agentic coding made it tractable. Meeting Reminder existed because I have ADHD and miss meetings, not because I had a Swift architecture worked out for it. ProPresenter Lyric Export existed because a small group of church tech volunteers needed it on a Sunday morning, not because I’d been holding out for the right TypeScript abstraction. None of those projects were waiting on the right idea. They were waiting on the cost of building dropping to something I could actually afford to spend.

What Claude does is collapse that cost. It is not the source of the taste, the scope decisions, the UX calls, or the choice of whether the feature is worth building at all. Those were mine. They were always mine. They would have stayed in my head forever if the construction cost hadn’t dropped close to zero.

The right reading of yesterday’s table isn’t Adam has ideas, Claude has the skills. It’s: same ideas as before, time constraint removed.

That’s the difference between AI is the smart one and AI is leverage. The first framing is wrong, and it’s wrong in a way I should be careful about, because it’s the framing that makes people both more frightened of these tools than they need to be (the machine is taking over) and more credulous of them than they ought to be (the machine knows best). Neither is true. The machine is fast. That’s the part of the story that actually changed.

The constraint that’s been removed is time. Not ability.

Thanks for the push.