This is an I-Heart podcast. Guaranteed human.

Callsum Media.

Hello and welcome to this week's Better Offline Monologue. I'm your host, Ed Zittron.

Better Offline.

And this has been a tough week. I'm not going to lie. Monday, we recorded a wonderful episode with Victoria's Song and Ashman Rodriguez, which got lost due to a technical fault. And then I recorded this entire monologue, which then got lost to a completely different computer, different room, different place, technical error.

I love making this show for all of you. It's been kind of a pisser, but it's important we get

on top of this goddamn subject. So check the episode notes, buy a challenge coin, read the newsletter.

There's a premium version unrelated to this show. I'd love if you'd subscribe, and if you don't,

I won't feel anything. Don't worry about it, but let's get to it.

Last week, Open AI, launched GPT5, a new flagship model of some sort that's allegedly

better at coding and writing, but in reality is much more of the same.

Another model is intermittently better at benchmarks built specifically for large language

models because they can't do actual work.

The Wall Street Journal reported late last year that it took multiple half-billion-dollar training

runs to get GPT-5 off the ground, and Altman himself said in the podcast with Theo Von,

of all people, that GPT-5 scared him and made him say, what have we done?

And that's a good bloody question, Sammy.

According to OpenAI,

GPT-5 is a unified system with a smart, efficient model that answers most questions, a deeper

reasoning model and a real-time router that quickly decides which model to use, based on conversation

type, complexity, tool needs, and your explicit intent. I read all of that out because I wanted

you to hear how convoluted GPT5 is and how much effort

...