Tertulia.

It happens in a podcast aside, a late-night post, a critic's column, a scientist's offhand recommendation. A title gets mentioned. Then mentioned again. Somewhere in that scatter of enthusiasm is the exact book you'd love - if only you could hear it over the noise.

Most of the internet loses that signal. Tertulia is built to catch it. Open the app and it hands you five books a day, chosen not because a warehouse has too many copies, but because people who think like you can't stop talking about them. It is, in the most literal sense, a bookstore that pays attention.

The name is a borrowed one. A tertulia is the old Spanish tradition of gathering in a cafe or bar to talk - about art, politics, and, gloriously, books. Sebastian Cwilich and his co-founders took that centuries-old idea, wired it to machine learning, and pointed it at the endless conversation already happening online. The result is less an algorithmic feed and more an overheard conversation you're glad you walked into.

Sebastian Cwilich had done this before. As a co-founder of Artsy, he helped build a discovery engine for a famously opaque market - fine art - where the good stuff hides behind gallery doors and insider whispers. Books, it turns out, have a similar problem in reverse: too much noise, not enough signal. Millions of titles, endless opinions, and a discovery experience that mostly amounts to a bestseller wall and an algorithm nudging you toward what you already bought.

So Cwilich reunited with Robert Lenne, Artsy's longtime head of product and design, and brought in Lynda Hammes, the former publisher of Foreign Affairs, who knew the industry from the inside. Their bet was simple to state and hard to build: the best recommendation engine in the world is the conversation that already exists. You just have to teach a machine to eavesdrop on it without embarrassing itself.

That embarrassment is real. Teaching software to tell a book conversation from a movie with the same name, or a person named after a novel, is genuinely thorny. Tertulia's answer was a hybrid - machine learning to find and classify mentions at scale, human curators to keep the taste intact. The technical brain of the operation lives in Buenos Aires; leadership runs from New York and Sweden. A small, distributed company for a very large firehose of chatter.

Then came the twist that separates Tertulia from the pack. Rather than treat customers as a funnel to be optimized, it invited them to own the place. The community co-op - $25 a year - hands readers actual ownership units, growing with every purchase. It's an old cooperative idea dressed in startup clothes, and it reframes the whole relationship. You're not a user being monetized. You're a shareholder who happens to love novels.

To make any of it work, Tertulia needed inventory - and got it. A partnership with Ingram Content Group opened the door to roughly 15 million titles, with fulfillment handled behind the scenes. John Ingram himself came on as an investor and board member. AlleyCorp led the $6 million raise, its founder Kevin Ryan taking a board seat, alongside a cast that reads like a media hall of fame: the founders of MTV, EarthLink and Shutterstock.

Return to that scattered conversation - the podcast aside, the late-night post, the critic's column. It hasn't stopped. It never stops. What's changed is that the noise now has a destination. A mention becomes a signal; a signal becomes one of your five picks tomorrow morning; a pick becomes a book on your doorstep, shipped free, bought from a store you partly own.

That's the quiet radicalism of Tertulia. It didn't invent the love of books, or the habit of recommending them - people have gathered in cafes to do exactly that for centuries. It just built the room big enough to hold the whole internet's version of that gathering, and then handed the guests a key. The tertulia was always a conversation among equals. This one comes with a shopping cart and a share of the till.