Tom Tomaszewski uses Codex to build freddy.coach, which connects people’s health and fitness data to AI, and ChatGPT Work to handle company paperwork. He says the business is already breaking even, with a two-person team serving more than 5,000 users around the world.
Tomaszewski, 37, grew up between two lakes in rural Poland, and now lives in Zurich with his family. He built his career by leading sales teams at robotics and battery-technology companies, not writing code. When AI made it possible for him to build software himself, he left to work on freddy full-time.
His interest in longevity grew out of becoming a father. Running, which he had never liked before, gave him personal time. He wanted to stay healthy enough to see his children grow up and have children of their own. Using AI to research the relationship between VO2 max, a measure of aerobic fitness, and longevity led him toward triathlons. Swimming was familiar from his childhood; cycling was new. A daily ChatGPT routine searched local used-bike listings and helped him find a bike that a knowledgeable friend endorsed as a great deal.
Training created a problem he wanted to solve: ChatGPT already knew a lot about him, but his latest health data was fragmented and slow to retrieve. “I want to give it fresh data,” he says. He built freddy to bring sleep, recovery and activity records into that conversation, along with years of history. It connects to wearables, rings, and even glucose monitors, through sources like Garmin, WHOOP, Oura and Apple Health, so people can ask questions about their own sleep, training and recovery in plain English. He says early users asked why their sleep had worsened and whether they had recovered enough to train hard.
Codex is where he spends most of his working day, developing the product and handling customer-support cases. He worked alone until August, when the product “outgrew me.” Then he brought in the co-founder of his former startup, a friend whose data-architecture experience now helps freddy process the huge volume of data that wearables produce. Devices can log several readings a second, producing hundreds of thousands of records per user each month. Together, they build for web, iOS and Android.
He uses ChatGPT Work for the paperwork of running a company, and to organize training and family trips. Without Work, he estimates, planning their frequent Swiss hiking and camping weekends would take two or three evenings. Work researches destinations and plans around his children and training schedule.
In ChatGPT’s voice conversations, he and his daughter explore questions about space and dinosaurs. His AI coaching routine helps plan daily training and races: at his latest triathlon, he says it helped him cut 15 minutes from the previous year’s time.
He says freddy has grown through word of mouth and product launches on Reddit, without paid advertising. AI lets the team keep up with support while shipping improvements every day. “We measure time in minutes and hours, and not in weeks and months,” he says.
