He built his first website at 12. Now his software decides which video plays next on some of the largest publishers on the open web.
Tom Pachys runs EX.CO, a video technology and monetization platform that sits, mostly invisibly, underneath a lot of the video you watch on the open web. When you land on a Hearst title, or Time, or Nasdaq, or a CBS Interactive property, and a video loads and an ad plays and another video queues up behind it, there is a decent chance EX.CO chose that sequence and captured that dollar.
This is the part of the internet nobody roots for and everybody depends on. Publishers make content, advertisers want attention, and in between sits a long, expensive chain of intermediaries taking a cut at every hop. Pachys's whole pitch is that the chain is too long, the cut is too big, and machine learning can fix both. He calls it supply-path optimization, which is an ad-tech phrase engineered to make eyes glaze over, and he has made it the entire mission of his company.
Machine learning should be used to solve problems that are impossible to solve.
It is the kind of line that sounds like a platitude until you notice what it excludes. He is not saying machine learning should power a slightly better dashboard, or auto-generate a headline, or write a press release about machine learning. He is saying: point it at the problems humans genuinely cannot crack by hand. In his case that means deciding, in real time, which video to show which reader at which second to earn the most money without wrecking the experience. That is a problem with too many variables and too little time for a person to solve, which, by his rule, makes it exactly the right problem for a machine.
Pachys was 12 when he built his first website. This is the sort of biographical detail that founders love to cite and that usually means very little, except that in his case the line never really breaks. He kept building. Before the funding rounds and the Disney money and the New York office, he served in an elite technology intelligence unit of the Israeli Defense Forces, which is where a striking number of Israeli software founders learn to ship under pressure with high stakes and no excuses.
He came out of the army into the video business early. He was among the first employees at SundaySky, a real-time video experience platform, where he rose to Director of Professional Services and Products - a title that means he was the person making sure the technology actually worked for the customers paying for it. Then, briefly in 2011, he co-founded and ran a machine-learning startup called Whimado. The tenure was short. The interest in machine learning was not; it would resurface a decade later as the core of everything EX.CO does.
In 2012, in Tel Aviv, Pachys co-founded Playbuzz with Shaul Olmert, son of former Israeli prime minister Ehud Olmert. Playbuzz became a phenomenon: the engine behind the quizzes, polls, lists, and "which character are you" widgets that colonized the internet in the mid-2010s. It was fun, viral, and, like most viral consumer plays, hard to turn into durable revenue. So Pachys did the unglamorous thing. In 2019 he led a full rebrand, retiring the Playbuzz name in favor of EX.CO, short for "the experience company," and repointed the whole operation from consumer engagement toys toward the far less flashy business of helping publishers make money from video.
Publishers never had access to such advantages without the likes of EX.CO.
Killing a beloved brand to build a boring one underneath it is the rare startup move that looks like a downgrade and is actually the point. Playbuzz was famous. EX.CO is infrastructure. Infrastructure is where the money and the moat live.
The clearest window into how Pachys thinks is the Bibblio acquisition. In 2022, EX.CO bought Bibblio, an AI and machine-learning company built around content recommendation. Most acquisitions of this kind end with the product quietly deprecated. Pachys did the opposite: he wired Bibblio's recommendation engine into large language models and made it a headline feature. The claim that came out the other side is specific enough to be checkable - contextually relevant videos, he says, lifted visitor dwell time by nearly 50 percent, and the recommendation engine drove publisher revenue up by as much as 17 percent.
"It's not only context that is important but also, recency of the article, popular interest, semantic analysis, safety filtering, and other metadata."
He reserves machine learning for decisions with too many variables and too little time for a human to make well - which video, which reader, which second.
"Our biggest focus is supply path optimization, making the value chain more efficient" - fewer intermediaries between publisher and ad dollar.
The through-line from Whimado in 2011 to the LLM engine in 2024 is not a coincidence. Pachys has been circling the same idea for his whole career: that the video business is drowning in decisions no human can make fast enough, and that the company which automates those decisions well captures the value everyone else is leaking. EX.CO is that idea, finally at scale, with Disney's money behind it.
Machine learning should be used to solve problems that are impossible to solve.
Publishers never had access to such advantages without the likes of EX.CO.
Contextually relevant videos have improved visitor dwell time by nearly 50% on average.
Our biggest focus is supply path optimization, making the value chain more efficient.
The open web can still out-monetize the walled gardens - if publishers get the right tools.
That is the bet, stripped to one sentence. The walled gardens - the big social and search platforms - won the attention economy partly on technology publishers never had. Pachys's wager is that with the right video player, the right recommendation engine, and a shorter supply chain, an ordinary publisher can compete on economics. Whether he is right is an open question. That he has spent more than a decade and two company names chasing it is not.
Tom Pachys is the co-founder and CEO of EX.CO, a Disney-backed video technology and monetization platform used by publishers including Hearst, CBSi, Time, Nasdaq, and VICE. He started building websites at 12, served in an elite Israeli intelligence-technology unit, was an early leader at SundaySky, and co-founded the machine-learning startup Whimado before co-founding Playbuzz in 2012, which he rebranded into EX.CO in 2019. He now leads roughly 160 people worldwide, betting that AI and supply-path optimization can rescue the economics of online video for the open web.
Last updated: