Notes from the working desk.
Long-form posts on product strategy at scale, the org-design problems that show up in AdTech, and the AI questions founders are actually asking me. Written by an operator, for operators. New pieces land roughly every other week.
The AI You're Paying For Is Cheaper Than It Should Be. That Won't Last.
The frontier AI we're all paying for is almost certainly priced below what it costs to run. Here's why that won't last, and why learning the alternative now is the calm move.
What "Local AI" Actually Means (It's More Like Owning a Car Than Calling a Taxi)
Running AI on your own computer sounds either trivial or impossible depending on who you ask. It's neither. Here's the plain-English version of what local AI is, why a gaming PC is often enough, and what you're not signing up for.
Why Bother? Give Your Local AI a Job: Meet the AI Butler
A local chatbot is just a more private version of what you already have. Wrap a model in an agent and it becomes a butler: one that remembers you, acts on your behalf, and answers to nobody else.
Picking a Model That Fits Your Machine (and Letting Claude Be Your Guide)
The biggest model that fits on your computer isn't always the one you should run. Here's the hardware fork (PC card vs. unified-memory Mac), the patient-vs-snappy trade-off, and why the frontier AI you already pay for is the best setup guide you can get.
The Weekend Build: Getting Your First Local Model Running
The actual build, the way I'd explain it to a friend over coffee. Skip the click-by-click and keep a frontier chatbot open as your patient setup guide. Two hours, one coffee, and a model running on your own machine.
Living With Local AI: What It's Great At, What It Isn't, and Why Now
After living with a local model alongside my frontier subscriptions, here's where it genuinely earns its place, where it doesn't, and why I think the weekend was time well spent, especially right now.
The Weekend a Model Got Banned (and Why I'm Glad Mine Lives Under My Desk)
Anthropic shipped Claude Fable 5 on June 9. Four days later it was gone. A weekend benchmark of two open models against the cloud tiers people actually reach for, and the case for getting serious about local.