I recently read the transcript of an entire conference day, from start to finish. Three people on stage, three talks, and all three said the same thing without saying it out loud.
One had built an app over the weekend without ever having written a line of code. A developer would have quoted him two to three months. The second said his backlog was filling up with 200 good ideas and building is no longer the problem. The problem is deciding which of them should exist. The third said AI only speeds up the wiring that's already there. It doesn't replace it.
Everyone talked about how fast you can build today. Nobody talked about who chooses what's been built in the end.
When building costs almost nothing
What used to seem expensive and defensible becomes a building block. A single developer builds an open-source alternative in a week to a product others needed years for. And according to an analysis cited by martech.org, around 63 percent of the people building with vibe coding today aren't developers. They build what they used to buy.
That sounds like the end of software as we know it. It's not quite that simple.
Vibe coding lowers the cost of building something. But not the cost of being wrong.
That's how the investor Forgepoint puts it, and I think it hits the mark. Whoever builds their own tool also owns the maintenance, the bugs and the responsibility for it. Most people don't want to build software at all. They want something to be done.
Nobody buys the hammer
That's not a new thought. Nobody ever wanted the hammer. People wanted the hole in the wall and someone who answers for it when the wall crumbles.
A 2024 study by Washington State University fits surprisingly well: when a product description says “artificial intelligence”, purchase intent drops for many people, because emotional trust drops. People don't buy technology. They buy less pain.
The buyer is changing
At the same time, something is moving in between providers and people. In its 2026 Hype Cycle for Digital Commerce, Gartner prominently placed terms like “agentic buying” and “answer engine optimization”. Protocols are emerging through which AI agents can pay. The future being described: an agent searches, compares and buys.
Almost all the numbers on this, however, come from companies that profit from exactly this future. That doesn't make them wrong. But I read them like a weather forecast from people who sell umbrellas. And honestly: nobody knows which standard will win.
What remains
- Building is the entry ticket, not the edge.
- Don't build every idea the AI delivers. Decide what's worth existing. That's strategy.
- Trust and attitude can't be copied, not even over a weekend.
- Don't rebuild on a hunch. First clarify what you offer and what you stand for.
A question to take with you: what about your offer can't be rebuilt over a weekend? If nothing comes to mind, that's the trail you should follow.