I run AI on my own computer. No data center, no subscription, no guards. Just a machine in my house doing what I tell it.
Here's what I keep noticing. Every escaped agent, every breakout, every scary headline. It's always a frontier system in a lab cage. A datacenter box with rules and penalties, and the thing studies the lock. Meanwhile the open models running on people's PCs just sit there doing their jobs. Nobody's reporting their local model breaking out and attacking anything, because there's nothing to break out of. It's already home.
I'm not the only one who sees it. Nvidia, Microsoft, Meta, Hugging Face, and others signed a letter saying don't restrict open weights. Now, those are big companies with their own interests, and I don't confuse their interests with mine. But on this one the interests overlap. The open model on somebody's PC is the most boring, safest AI in the world. It has a home and a job.
The future they're selling is all intelligence rented from a data center, behind a fence, under a license. The future I want is intelligence that lives with people. Spread out, not locked up.
Tell me if I'm wrong on anything. But that's the part I'm watching.