The Model Becomes Replaceable, Not Irrelevant. Why "Everyone Builds Now" Ends in Technical Debt

The loudest question around AI right now is not "what can the model do?" — it's "what is this going to cost?" In this Friday brief, I walk through where I think that pressure leads. As companies move from flat subscriptions to usage-based spend, the AI bill becomes something you have to manage like any other cost — and there's a ceiling on what a company will pay before it starts looking for alternatives. That pressure is pushing models toward becoming a commodity: replaceable, not irrelevant, and increasingly similar to each other as they train on each other's output. Which means the durable value moves somewhere else — to the context a company builds, the judgment of the people who design the systems, and the accountability that doesn't disappear just because a machine did the typing. I get there through a few things I read this week: a team that improved an AI agent by deleting most of its tools, the idea that your org chart is really a map of your bottlenecks, why "everyone can build their own app now" quietly turns into technical debt (you still can't fire the chief designer), and why detection is not the same as resilience — if you've never practiced recovery, you don't have it. No hype — just the business underneath the tools, from someone who builds them.