Descriptions:
Matthew Berman uses MyFitnessPal’s acquisition of Cal AI — reported as an eight-figure deal after nearly a year of negotiations — as a lens for examining how quickly AI coding tools are commoditizing entire app categories. Cal AI’s core feature, pointing a phone camera at food to estimate calorie and macronutrient content, can now be replicated by a solo developer in under an hour using off-the-shelf AI agents.
To make the point concrete, Berman builds a working calorie tracker live on screen between 11:18 and 11:35 AM — 17 minutes — using the Cursor agent CLI running inside OpenClaw. The resulting tool uses Gemini Vision to identify food items, estimate portions, and return per-item and total figures for calories, protein, carbs, fat, and fiber, stored in encrypted SQLite with daily target tracking. A live test with an In-N-Out burger photo returns 744 calories against In-N-Out’s published figure of 610 — close enough to be practically useful, and comparable to Cal AI’s own accuracy.
Berman traces MyFitnessPal’s history: founded 2005, acquired by Under Armour for $475M in 2015, sold to private equity for $345M in 2020 — a $130M loss — and now acquiring Cal AI as a growth move. He argues non-technical ownership at each stage prevented MyFitnessPal from adapting, and that the acquisition reflects a broader pattern of incumbents misreading how AI agents will absorb standalone app functionality. Cal AI’s founders — seven employees including CEO Zack Yadagari — are credited with exiting at an optimal moment.
📺 Source: Matthew Berman · Published March 03, 2026
🏷️ Format: Hands On Build
