A weekly show about AI that runs on your own hardware.
I'm Frank Buchholz from Cadence 3 Momentum, joined by my Surface colleagues Jacob Rhoades, Robert Henry, Neil Misak and Chauncey Larsen. They're still at Microsoft. I'm the odd man out these days, but it's good to get the crew together every week.
We speak for ourselves, not our employers.
localhost:0006 · Sep 26, 2026 · 44 min
Ever nod along to an AI conversation and quietly have no idea what a runtime or a parameter actually is? This episode is for you. We go back to the studs and build local AI from the silicon up, one layer at a time, with zero jargon left unexplained. If you can follow a kitchen, you can follow this.
localhost:0005 · Sep 20, 2026 · 40 min
Five guys, five machines, and one question: what actually works? No debate this week, no vote. We opened up our laptops and showed each other what we are actually running, what broke, and what turned out to be worth the trouble.
localhost:0004 · Sep 14, 2026 · 41 min
An escape hatch is the thing that lets you leave. Open weight models were supposed to be that hatch for AI: download them, run them yourself, walk away whenever you want. Then NVIDIA bought Hugging Face, the place everyone downloads them from. Is the hatch still open, or did the company you would escape from just buy the exit?
localhost:0003 · Sep 6, 2026 · 45 min · Hot seat: Frank
Last week we argued the $25 token was dying. This week: that is exactly when the hardware boom starts. If intelligence keeps getting cheaper, we invent more things to do with it, and every enterprise gets a new question: which intelligence do you rent, and which do you own?
localhost:0002 · Aug 30, 2026 · 38 min · Hot seat: Robert
If being smart gets cheap, what is left to charge for? Frontier models are getting dramatically cheaper, the models you can run on your own hardware are getting dramatically better, and Robert argues that within 18 months the only thing frontier labs can really sell is speed.
localhost:0001 · Aug 21, 2026 · 42 min · Hot seat: Frank
Meta said its new open model beats Google's and Alibaba's. So we benchmarked all three on a used $700 RTX 3090, then argued about what open really means. Plus a job interview for five AI models, with a winner nobody was talking about.
Nothing loads from Spotify or YouTube until you press a button on this page. Same first-party rules as the rest of the site.
There's no place like 127.0.0.1.