Booking.com Has a Research Blog and I've Been Missing It

There's a specific humiliation in discovering that a company you've been using for years to find overpriced hotel rooms — a company whose UI you have strong opinions about, mostly negative — has been quietly running a substantial AI research operation the entire time.

booking.ai. It's just sitting there.

The mental model you carry around for Booking.com is: big search box, suspicious "only 2 rooms left!" countdown timers, dark patterns engineered by someone with a philosophy degree and a conversion rate target. That's the company. You're sure of it.

And then you find out they have a research blog full of papers on reinforcement learning and ranking systems and demand forecasting at a scale that most AI shops will never touch — because they have hundreds of millions of users doing real things with real money in real time, which is the kind of messy ground truth that makes academic datasets look like dollhouses.

The thing about consumer internet companies is that the ones you use every day are running ML experiments on a scale that pure research labs can only approximate. Booking has been doing this in the background, publishing it, and I have been — apparently — looking the other way.

No real conclusion here. Just that the map I keep in my head of "who is doing interesting ML work" has another gap in it, and the gap is shaped exactly like a hotel search bar.