🇺🇸 Zach Yadegari

Zach Yadegari's first exit came at sixteen. The product was Totally Science, a website hosting 'unblocked' browser games — the category of site that exists specifically to get around the content filters school districts install on student laptops. It is not a serious technology company, and Yadegari has never pretended otherwise. What it was, was a real business with real traffic, real infrastructure problems, real user acquisition and a real buyer. He learned distribution before he learned engineering, which turned out to be the more valuable sequence.

Cal AI came next, built with a co-founder while both were in high school. The premise was mechanical: point a phone camera at a plate of food and get a calorie estimate back. The insight was that image models had crossed the threshold where this actually worked well enough to be useful, and that nobody had shipped it in a form ordinary people would use daily. Manual calorie logging — the mechanic MyFitnessPal was built on — has an enormous drop-off rate precisely because typing every meal into a database is tedious. Photograph-and-forget removes the friction that kills retention.

The growth was extraordinary by any standard and almost unprecedented for a product built by teenagers. Cal AI passed fifteen million downloads and crossed $30 million in annual recurring revenue in under two years, with revenue over the trailing twelve months reported at approximately $40 million and a projected $50 million for 2026. Those numbers were achieved by a team of seven people. For comparison, most venture-funded consumer subscription apps that reach $30 million in annual revenue employ several hundred.

The distribution strategy was as important as the technology, and it was the part that came from Totally Science. Cal AI grew primarily through short-form video — TikTok, Reels, Shorts — where a two-second demonstration of photographing a burrito and getting a number back is a complete advertisement. This is a channel that established fitness companies with large marketing budgets have repeatedly failed to master, and a channel that a teenager building for other teenagers understands natively.

MyFitnessPal's decision to acquire rather than replicate is the most informative part of the story. The company had the data, the brand, the users and the engineering capacity to build a photo-based logger. It bought one instead. The deal closed in December 2025 and was announced publicly in March 2026, and Yadegari along with the entire seven-person Cal AI team were retained. Incumbents buy when the challenger's growth rate, rather than its technology, is the asset they cannot construct internally.

Yadegari was nineteen at the time of the announcement and a freshman at the University of Miami — a detail that inverts the standard founder narrative. The dropout myth is so entrenched in technology that staying enrolled reads as the contrarian move. He has been open about the calculation: the company was acquired and the team retained, which meant the operational burden that usually forces the choice had been resolved by the transaction itself.

The scepticism the story attracts is worth stating directly. AI photo calorie estimation is approximate, and the nutrition-science community has been vocal that portion estimation from a single image carries meaningful error bars. Yadegari has never claimed clinical precision. The product's argument is behavioural rather than analytical: a rough number logged every day beats a precise number logged for a week and abandoned. Whether that holds up under long-term study is unresolved, and it is the substantive open question about the entire category.

What is not in dispute is the commercial result. Two companies built and sold before the age of twenty; fifteen million downloads; a revenue run-rate that would place Cal AI in the top tier of consumer subscription apps; and an acquisition by the category's incumbent. Yadegari has said he is aiming higher next. The pattern established so far — find a mechanic that removes friction, distribute it through a channel incumbents cannot use, and stay small — is a repeatable one.