Gears Within Gears

Introducing Prospero: Superpowers for writing

Thu, 30 Apr 2026 10:00:00 -0400

No man but a blockhead ever wrote, except for money. — Samuel Johnson

I like the act of writing, but often find actually publishing paralyzing. I’ll sit with a piece for weeks, hem and haw over it and chew on it and maybe pass a draft around, but by the time I’m done I’ll have lost whatever conviction I started with. A fair number of pieces never finish, and the drafts folder keeps growing. Consequently, I tend to think of myself as a better editor than an author; responding to a piece is less taxing than authoring one.

Software architecture after AI

Tue, 21 Apr 2026 06:00:00 -0500

One of the more useful definitions of software architecture comes from Building Evolutionary Architectures: architecture is definitionally the stuff that’s hard to change.¹ I’ve always found this definition to be the most honest framing available, to say nothing of the simplest. It doesn’t pretend architecture is about beauty or correctness or your resident architect’s favorite stalking-horse. It acknowledges that what makes a decision “architectural” is not its conceptual weight but its cost to reverse and its business impact. And “hard to change” has always been, at root, about wall-clock time: coordination cost, incident mitigation, cognitive load, handoff friction. Software architecture has always been a labor problem dressed up as a design problem.

AI isn't a value-add for software; software is a value-add for AI

Sat, 14 Mar 2026 10:00:00 -0400

When a company hits a ceiling on its core business, it often tries to grow demand. The Michelin Guide wasn’t about restaurants; it was about miles driven. Facebook’s internet.org wasn’t about altruism; it unlocked expansion by growing the size of the entire internet. The ceiling on the business was the medium, so they expanded it.

AI companies face the same structural problem: their constraint on inference revenue is the volume of tasks that flow through models. From where they sit, every piece of software that still runs without model calls is unrealized market.

Accumulated ignorance at scale

Fri, 06 Mar 2026 12:00:00 -0500

We have entered a period in which AI is able to author more code than is reviewable by humans. Many teams are already past the point where each line can realistically be scrutinized; the volume of code being generated is increasing faster than our capacity to inspect it. Every AI-assisted coding session produces choices: naming decisions, structural decisions, abstraction decisions, dependency decisions. Some of these are good, some less so, and the ratio of decisions being made to decisions being reviewed by a human is growing.