We haven't hosted a fitness founder yet, so AI in fitness on Built This Week comes from Sam's own routine and from marketing HipTrain, one of our businesses, an app for one-on-one personal training. Sam replaced his calorie app with a ChatGPT thread, then built a personal trainer that reacts to every set. The common thread: AI takes over the logging, the part Sam kept giving up on, and answers right away. Treat this page as first-hand notes, not a verdict: the trainer was still in beta when Sam demoed it.
If you can imagine a trainer after every rep is like, you actually struggled there on that 10 reps.
— Sam Nadler, Episode 30
How companies we talked to use AI in fitness
Juma: a month of fitness content from one prompt
In Episode 35, Iliya Valchanov of Juma, an AI agent for marketing tasks, picked HipTrain for his demo. He described it as an app that helps with personal one-on-one training, and decided the website was fine but social media could use help. One prompt produced an April calendar for LinkedIn, X and Instagram with pillars such as wellness, fitness education and motivation, then an Instagram carousel. Iliya said a project keeps memory of what it posted, and with the new Instagram connector it should learn what performed.
HipTrain: AI ad creative for a training app
In Episode 40, Sam asked Claude Design for Meta ad concepts targeting women between 50 and 70, which he called one of HipTrain's prime demographics, in a cheeky, self-aware tone. It returned three campaigns, including "The grandkids think I'm chill," each with an image prompt and a Veo 3 prompt. He ran one image prompt through Nano Banana, handed the image back to Claude Design to finish the ad, and shared the deck with the marketing team for comments.
What we built
- ChatGPT as a nutritionist. Not an app, just a ChatGPT project. Sam had cycled on and off MyFitnessPal because measuring 1.5 cups of rice or five ounces of chicken got tedious. Now he sends a photo or a voice note, gets calories and macros back (one lunch came to 460 calories and 42 grams of protein), plus tips for what to eat later in the day. It costs him nothing extra, and the project keeps his history. Later, he said pairing it with Jordan's Ryz Chef recipe app would be a powerful combination. (Episode 4)
- Hollywood transformation trainer: a mobile web app for aesthetic upper-body muscle gain and progressive overload, "for my premiere as a Hollywood star," Sam joked. He gave Claude Cowork a folder with the PRD, typed "build this," and picked React, Supabase and a real LLM API for the training responses. It was not finished on air.
- Rep-by-rep AI personal trainer: settings for workout length, goal (bigger vs. stronger), training days, body focus, experience, equipment and limitations. When Sam logged a set as a struggle, a fatigue check stepped the weight down from 20 to 7.5 and cut the reps, and an energy check came up between exercises. He polished it in OpenAI Codex with the extra-high reasoning model, where runs took five to ten minutes.
I just voice noted in every time I eat something, or I just take a picture of the product. Doesn't even have to be the barcode.
— Sam Nadler, Episode 4
What's still hard
- It won't say no. Sam's AI nutritionist never refuses. Ask about ice cream and it finds a way to say yes.
- Logging still has friction. The trainer exists because typing every set into a ChatGPT thread at the gym got annoying. Sam wanted the interface to be super simple.
- Deep reasoning is slow to build with. Sam found Codex's extra-high model great at finding bugs but a longer, less engaging way to work than Claude Code. His advice: a faster, lighter model for quick iterations, the deep one for polishing toward production.
- Untested under a barbell. At the time of Episode 30, Sam was still using the ChatGPT thread and planned to start testing the app the following week.
Jordan's take is that this is personal software: instead of a fitness app or a spreadsheet, it's now easy to build the exact tool you want for yourself. Jordan's DNA and blood work dashboard from the same episode is covered on our AI in healthcare page.
Episodes on AI in fitness
- Episode 4: Sam's ChatGPT nutrition tracker (July 2025).
- Episode 29: the Hollywood transformation trainer started live in Claude Cowork (January 2026).
- Episode 30: the rep-by-rep AI personal trainer, built with Codex (February 2026).
- Episode 35: Juma builds a social calendar for HipTrain (March 2026).
- Episode 40: HipTrain ad concepts in Claude Design (April 2026).
FAQ
How is AI used in fitness?
On our show, Sam tracks calories and macros by sending photos and voice notes to a ChatGPT thread, and he built an AI personal trainer that adjusts weight and reps after every logged set. We also use AI to draft ads and social content for HipTrain, an app for one-on-one personal training.
Can ChatGPT replace MyFitnessPal?
It did for Sam. He snaps a photo or records a voice note for each meal, and ChatGPT returns calories, protein, carbs and fat, plus ideas for what to eat later to hit his targets. The catch: it never tells him no when he asks for ice cream.
Can you build your own AI personal trainer?
Sam did. He started a Hollywood transformation trainer live in Claude Cowork from a PRD, then showed a rep-by-rep trainer he polished in the Codex desktop app with GPT-5.2 Codex at extra-high reasoning. It was still in beta at the time.
How can a fitness brand use AI for marketing?
For HipTrain, Sam had Claude Design draft three cheeky Meta ad campaigns aimed at women aged 50 to 70, and Juma CEO Iliya Valchanov generated an April social calendar with pillars like fitness education and motivation from one prompt.