Will AI Eat Software?
There’s a narrative building that I think is mostly wrong, and it’s worth saying so plainly: AI is not going to eat software. It’s going to change it, reshape it, compress certain parts of it - but the idea that software companies are heading for extinction because someone vibed a to-do app into existence over a weekend is a category error dressed up as a prediction.
I keep seeing the same argument recycled. Vibe-coding is here. Anyone can build an app now. Software companies are dead. The sell-off is the beginning of the end. I understand the excitement. I’ve used these tools. They are genuinely impressive. But impressive is not the same as sufficient.
What a Vibed App Doesn’t Come With
Here’s the thing about software that the vibe-coding narrative conveniently forgets: software companies don’t just write code. They provide a service. They have obligations (contractual, regulatory, reputational) to the people who rely on them. An SLA. A security team. A compliance framework. A support desk that answers the phone when something breaks at 2am on a Friday.
A vibed app comes with none of that. It comes with a prompt and a prayer.
I’m not dismissing what vibe-coding enables. For prototyping, for personal tools, for getting from idea to something tangible, it’s a step-change. A cottage industry has sprung up around it and for good reason. But there is a gap between building something and running something, and that gap is where software companies live. It’s where trust lives. It’s where accountability lives. And I don’t see AI closing that gap anytime soon.
The Correction Isn’t the Collapse
Are software companies overpriced? Probably. Years of cheap capital inflated SaaS valuations beyond what the underlying businesses could justify, and the market is doing what markets do when the tide goes out.
But a correction is not a collapse. The fact that multiples needed to come down doesn’t mean the ground beneath them has vanished. There’s a difference between “this was overvalued” and “this has no value,” and the current discourse seems determined to blur it.
That’s not the point I’m making, though. Valuations are someone else’s problem.
The Circular Reference
The point I’m making is simpler, and I think it’s the one that matters: if AI eats software, it eats itself.
Think about what AI actually is. The large language models, the inference infrastructure, the training pipelines, the APIs, the orchestration layers, the monitoring, the deployment tooling — all of it is software. Built by software engineers. Maintained by software teams. Sold by software companies. The thing that’s supposedly killing the industry is the single biggest driver of software demand we’ve seen in a generation.
If you’ve ever watched a circular reference resolve in a spreadsheet, you know it doesn’t end in enlightenment. It ends in #REF!. The “AI eats software” thesis has the same structural problem - it devours its own premise.
What AI Actually Does to Software
What AI will do, what it’s already doing, is raise the floor. It compresses the time from idea to prototype. It automates the tedious. It makes it possible for more people to build more things, faster. That’s genuinely good, and I’m not interested in downplaying it.
But raising the floor is not the same as removing the ceiling. Someone still needs to be accountable when the system goes down. Someone still needs to think about governance, auditability, the boring but essential scaffolding that turns a clever demo into a service you can bet your business on. AI doesn’t eliminate that need. If anything, it intensifies it because now there’s more software, built faster, by more people, with less understanding of what’s underneath.
Software companies that get this - that understand their value isn’t in lines of code but in the trust, reliability, and accountability wrapped around those lines - will be fine. Better than fine. The ones that were only ever selling code? They were already in trouble before AI showed up.
Don’t be too bearish on software. The reports of its death are, as usual, greatly exaggerated.