Episode 7 · Greg Mushen · August 21, 2026 · 52 min
YouTube · Spotify · Apple Podcasts
About this episode
Greg Mushen built a personal AI agent that tracks meals, movement, and health data without making him live inside tracking apps. Instead of tapping through forms all day, he can message Hermes Agent in Telegram, send a photo or barcode, and let scheduled check-ins keep the record current.
In this episode, Michael and Greg break down what makes that system work—and where personal AI becomes much more than a chatbot.
What you’ll learn
- How Greg replaced manual food and activity logging with a conversational AI workflow
- How Hermes Agent uses Telegram, scheduled check-ins, and nutrition data to reduce tracking friction
- Why making a health system easier to use can matter more than adding another dashboard
- How AI helped audit 54 cited research papers behind a daylight-saving-time claim
- How Greg runs Gemma locally on NVIDIA Jetson AGX Orin hardware for sensitive health-data work
- Why local Whisper transcription, embeddings, and classification can keep private data off third-party APIs
- How Claude Code and OpenAI Codex change the way Greg builds and reviews software
- Where autonomous agents are useful today—and where human judgment still belongs
Topics and resources
- Greg Mushen on X
- Hermes Agent
- Telegram
- Cronometer
- Pedometer++
- USDA FoodData Central
- Apple Health
- Google Gemma
- NVIDIA Jetson AGX Orin
- OpenAI Whisper
- Claude Code
- OpenAI Codex
More Already Here
Show home · All episodes · Previous: Episode 6 · Next: Episode 8