This is an I-Heart podcast.

Guaranteed human.

Quarzone Media.

Hi, my name's Ed Zittron and welcome to Better Offline.

This is also Jackass.

Better Online. Better Online. So you've just had a cheery two-part chuckle first about how generative AI made tank our markets in our economy. So I'm going to give you a lighter one, an episode about GPT-5,

which is a model from OpenAI, and why just under three years of hype have led to the software equivalent of the launch of St. Anger, except every time Lars Orick hit the snare drum, it cost them $55,000.

Now, if we look at the positive reviews, we see takes ranging from Simon Willison's tepid remark that GPT5 is just good at stuff. The semi-analysis is completely insane

statement that GPT5 is setting the stage for ad monetization and the OpenAI GBT, chat GPT super

app in a piece that makes several assertions about how the router that underpins GPT5 is somehow

the secret way the open AI will inject ads, which is just distinctly silly.

I'll get into this in the episode a little bit, but just with everything you're going to hear, you're going to realize that this is just someone just saying stuff.

Took four bylines to do that shit too. I'm also British. I'm going to say router as well, because I've been here

a while. Make fun of my voice if you really must. But with that out the way, here's a quote from

semi-analysis's coverage. Before the router, there was no way for a query to be distinguished, and after

the router, the first low-value query could be routed to a GPT5 mini model that can answer with zero

tool calls and no reasoning.

This likely means serving this user is approaching the cost of a search query.

This does not make any sense.

None of this makes it like, it's just a bunch of assumptions.

Why would this be the case?

The article also makes a lot of claims about the value of a question and how chat GPT could,

...