IM 890: 200 Donuts - How One Engineer Manages a Team of 24 AIs
Summary
In this episode, hosts Jeff Jarvis and Paris Martineau interview Mike Gennady, who runs a company of 14 AI employees to explore the potential of local, open-weight models. They discuss how hobbyists can use hardware like NVIDIA Sparks to run autonomous AI agents for tasks ranging from coding to education, effectively creating a "homebrew" computing environment. The conversation also covers OpenAI’s recent Dev Day announcements, including new pricing structures and the growing gap between frontier cloud models and high-performance local inference.
Topics discussed
Show intro and panel introductions
Mike Gennady joins: Local AI and hardware
Building a personal AI fleet and workflow
AI agents, memory systems, and personas
Small language models and local hardware
Sponsor: Framer AI website builder
Benito's AI puzzle and coding demo
Anthropic IPO prospectus and revenue risks
Google Gemini 4 and cybersecurity concerns
Sponsor: OutSystems AI agents
Jensen Huang, AI safety, and governance
Sponsor: AI endpoint security
Leo's trip to Angkor Wat and AI detox
Paris's medical checkup and oxygen therapy
Sponsor: Palo Alto Networks identity security
Fei-Fei Lee, AMD, and new AI labs
Anthropic, religion, and AI consciousness
AI pain perception and Milgram study
Dynamic pricing and fast food costs
Coffee, Mac Minis, and Cray supercomputers
Writing rituals and book royalties
Picks of the week: Movies and Fat Bear Week
Health updates and San Francisco observations
Food trends and Taco Bell compass
Discussion on the Elon Musk documentary
Book promotion and show outro
Listen ad-free on Castria