the localhost is live.

Two weeks ago I benchmarked three AI models on a used $700 graphics card to settle an argument. Turns out the argument was the real product.

So here it is: the localhost, a weekly podcast about AI that runs on your own hardware. Open weight models, NPUs, home labs, and the question underneath all of it: who controls intelligence, you or a data center?

The format is simple. Every week one of us takes the hot seat to defend a provocative claim. The other four try to take it apart. Then we all vote, and we keep score. In episode one I argued that openness is positional, not principled. The room went 3 to 1 with me. Robert remains unconvinced. That is the show working.

The five of us: me, plus Jacob Rhoades, Robert Henry, Neil Misak, and Chauncey Larsen. Four are Microsoft folks, opinions entirely their own, disclosures on air. I am independent. We have been having this exact conversation in group chats for years. We finally just hit record.

Why do it? Because local AI is having its moment and most of the coverage is press releases. We download the models, test the claims on our own machines, and publish what actually happened, including our own mistakes. Episode one has the full bake-off story, a job interview for five AI models, and one host's setup that runs 86 percent of his AI locally and has already saved him $1,600.

All the ways in

• One address, every player thelocalhost.show
• Spotify open.spotify.com/show/2xrxS4cpdK8Gk1dgfubbzI
• Apple Podcasts podcasts.apple.com/us/podcast/the-localhost
• YouTube, full video episodes youtube.com/@thelocalhost-show
• The receipts: episode one's full benchmark, raw traces included github.com/frankcx1/bakeoff

To say it once: if you listen and it is your kind of thing, a rating on Apple or Spotify genuinely helps a brand new show get found. And thank you for two weeks of encouragement on the bake-off posts. This is where it was all headed.

It runs a little long. First episodes do.

There's no place like 127.0.0.1.