I don’t know. That’s why I’m writing this.

I’ve been building and investing in software long enough to have a healthy distrust of clean narratives. And right now, the AI-is-eating-software narrative is everywhere — delivered with the kind of confidence that usually means someone stopped asking hard questions.

The claim goes like this: AI collapses the cost of building software, which destroys the moats, which kills margins, which ends SaaS as we know it. A solo founder can ship in weeks what used to take a team and a year. Why pay for a product when you can build your own? Why pay per seat when an agent just does the work?

Parts of that are clearly true. The production cost of software is collapsing. Narrow point-solution products — things that do one thing repetitively — are genuinely under threat. SaaS gross margins are compressing as inference costs land on the income statement in ways that traditional software costs never did.

But the story breaks down fast when you push on it. Distribution doesn’t get cheaper when code gets cheaper. Enterprise switching costs don’t evaporate. Twenty years of integrations, compliance certifications, and customer relationships don’t disappear because a new entrant can build the features faster. Salesforce isn’t worth what it’s worth because of its code.

So which is it? Both, partially. And that partial overlap is exactly where the interesting analysis lives — and where most commentary gives up and picks a side.

I’m not going to do that. This publication starts with a hypothesis — software loses its value because AI commoditizes its creation — and takes both sides of it seriously. What would have to be true for this to be right? What evidence would prove it wrong? When new data comes in, the hypothesis updates. No predetermined conclusion.

Weekly, I’ll look at a mechanism, a company, a data point — something concrete that moves the needle in one direction or the other. Pricing model shifts. Earnings calls where the numbers tell a different story than the press release. Structural changes in how software gets bought, sold, and replaced.

If the answer eventually becomes clear, I’ll say so. Until then — we follow the evidence.

— Chris

Sources & further reading