A friend once told me he wanted to vibe-code the next Airbnb. I understood the excitement. With AI and the right tools, listings, search, maps, bookings and payments can appear surprisingly quickly.

The surface can arrive before the business has really been understood.

But a marketplace is not only a checkout flow. Who holds the money before the stay happens? If a guest in South Africa books a place in Japan, how does the host get paid? If the guest charges back after the host has already been paid, who carries the loss?

Then the operational questions begin. What happens if the host cancels after flights are booked? If the property does not exist? If both sides of a dispute sound believable?

And someone still has to decide how to spot fake accounts, handle failed payouts, record who changed a refund, and recognise which bookings need human review.

This is where product work gets interesting. The screen is where the system becomes visible. Underneath it are policies, incentives, trust, cash timing, support tools, and the problem of finding enough good hosts before there are enough guests.

I love that AI can help people build faster. I use it every day. But speed creates a particular kind of pressure when the demo feels like proof before the hard questions have been asked.

The product is the whole system that keeps working when reality stops following the happy path.

Originally published on 29 May 2026. Read on jamesgodwin.me.