What AI Geeks Do for Fun: Mark Hinkle’s Multi-Model Playground

About this event

A new series where people who are deep into AI show what they actually run, build, and tinker with when nobody’s making them. First up: Mark Hinkle on multi-model workflows, homegrown agents, and the projects he vibe coded just to see if he could. So here’s the thing about AI talks. Most of them are either vendor demos or “10 prompts that will change your life.” This isn’t that. What AI Geeks Do for Fun is a new series where practitioners open up their own setups and show the unpolished, “I built this on a Saturday” side of working with AI. What’s on their machines, which models they trust for what, what broke, and what they can’t stop playing with. I’m going first, because it’s only fair to embarrass myself before I ask anyone else to. I’ll walk through: • My multi-model workflow. Why I don’t pick one model. How I route work across Claude, ChatGPT, Grok, and local open source models, and how they review each other’s work. • How I build agents. My running agent setup on a local Qwen model with OpenRouter as the fallback. What agents are good at, where they fall over, and why I think of them as entry-level employees who never get tired and occasionally make things up. • Mesh LLM. Pooling spare GPUs across machines into your own peer-to-peer inference cloud. Yes, a Mac Mini is involved. Stuff I vibe coded: • ConferenceOS: the conference management system I’m building so All Things AI can stop paying Eventbrite. • The Gibson: a harness for running fleets of agents. It’s named after exactly what you think it’s named after. • What I think is cool right now and what I’m keeping an eye on. Expect live demos, some things that won’t work on the first try, and plenty of time for questions. Bring your own setups and war stories. The best part of these nights is the hallway conversation. Who should come: Developers, builders, tinkerers, and curious business folks who want to see how AI gets used outside the slide decks. You don’t need to be an expert, but you should enjoy going down rabbit holes.