Lucky Gunasekara - the designer who taught the news to answer back.
Profile / Founder & CEO
Lucky Gunasekara
He once built a crawler to track disease outbreaks. Now he builds answer engines that pay the writers they quote.
He once built a crawler to track disease outbreaks. Now he builds answer engines that pay the writers they quote.
Lucky Gunasekara runs Miso.ai, an AI media lab that builds private answer engines for the people who make the news. The pitch is small and enormous at once: when a reader asks a question, the machine should answer in a sentence, name the sources it leaned on, and then quietly send those sources a royalty. He calls the idea generative royalties. The O'Reilly Radar headline he co-wrote in 2024 put it more bluntly - the R in RAG stands for Royalties.
That is the unusual part. Most of the AI conversation around publishing is about defense: block the bots, sue the scrapers, wall off the archive. Gunasekara works the other side of the equation. He thinks information wants to be liquid - remixed and repackaged in line with what a reader actually needs in the moment - and that the job is to make that flow without robbing the people who wrote the underlying words. Credit, consent, compensation. Those three words show up wherever he talks.
Miso partners with somewhere between 45 and 50-plus publishers and learning platforms worldwide. Its earliest believer was O'Reilly Media, where Miso plugged an answer engine into the learning platform and watched usage climb. The lesson stuck with him: readers reward frictionlessness. "You see a much higher rate of usage when you move away from high-friction search to very frictionless answers," he has said. The shelf full of books becomes a single question - what do you need right now?
He is, by training, an interaction designer, not a data scientist. That detail explains a lot. Miso's obsession is the feel of the thing - the moment a reader gets a clean answer - rather than the raw size of the model behind it. The plumbing is serious (an assembly line of LLM workers, each with a narrow job, reasoning about intent, researching, then grading their own answers) but the product is judged by the reader's experience, not the benchmark.
"I would be calling every publisher I know - hey, let's have dinner Thursday. It's an asteroid coming for all of us."
- Lucky Gunasekara, on AI and the news business
He studied Neurobiology and Behavior at Cornell, then headed to Stanford's School of Medicine. The white coat did not take - he says he dropped out in his 20s and went into AI instead.
As CIO at Metabiota he led R&D on epidemic surveillance, building a massive RSS crawler that pulled real-time signals from hundreds of news sources to spot disease before the headlines did.
At Cornell Tech's Small Data Lab he and co-founder Andy Hsieh chased private, personal AI - the opposite of the big-data gold rush. The crawler became a research engine. The research engine became Miso.
Static articles. Constrained search boxes. Newsletters you scroll past. Gunasekara thinks the next shift is not about which engine wins - it is about how information moves at all.
"Instead of having solid, static media like articles or constrained interfaces like search or newsletters, you get to a place where information is more remixed and repackaged directly in line with your interests."
The old promise was "look at all these books on the wall." The new one is "what do you need right now?" Lower the friction and people use it more - which is exactly what happened the first time Miso shipped answers into a publisher's platform.
Miso's monitoring of roughly 7,000 publishers turned up an uncomfortable picture of how AI crawlers behave - and how fast their success rates climb, robots.txt or not.
Figures reported by Miso.ai. Illustrative of the trend, not a precise measurement of any single publisher.
"The R in 'RAG' stands for 'Royalties.'"
"You see a much higher rate of usage when you move away from high-friction search to very frictionless answers."
"It's an asteroid coming for all of us."
"Information is more remixed and repackaged directly in line with your interests."
He came up through interaction design and UX research, which is why Miso argues about how an answer feels before it argues about model size.
The same crawler logic he built to catch disease outbreaks early became the seed of a tool for catching the right answer fast.
He frames AI as an asteroid for the news business - and then insists the move is to host a dinner, not to duck. Threat and opportunity in one breath.
A conversation with the Miso co-founder on what happens when the news starts answering back - answer engines, royalties, and the case for collaboration over panic.
Build private, privacy-first AI for publishers. Answer the reader in a sentence. Name the sources. Pay them. Repeat until the news business has a model that survives the asteroid.