SLHamlet

Yeah the author yanked the original version due to some liability concerns, edited it and reposted it with some new stuff:

https://bsky.app/profile/segyges.bsky.social/post/3myklta3iv...

lol yes

BTW this data is drawn from a US map of the entire US data center built-out, county by county:

https://pressaudit.org/blog/datacenters

The interactive map of the data center buildout she cites in the article is here:

https://pressaudit.org/blog/datacenters

I helped build it out / launch it, so happy to help answer questions about it.

Um, wow:

"Machine Gods, NPR’s new video podcast from former Hard Fork hosts Kevin Roose and Casey Newton, is receiving a $480,000 grant from the organization. Tarbell is also separately backing NPR’s Power and Influence desk, giving the public media outlet $250,000 to help expand its coverage of how AI intersects with politics, money, and power."

Also Tarbell helped Hard Fork get started, says Roose:

“Their funding has been very useful in helping us get started. It was a way for us to de-risk going out on our own. We’re feeling quite good about it and we’re grateful.”

I'm sure they'll closely report how non-open source LLMs are wildly unprofitable and unscalable... right?

Where the hyperscalers are landing, and where communities are pushing back, mapped against the per-capita income of the counties absorbing them.

Shit is really bananas! This is one of the rare cases where it's difficult for the mainstream media to accurately cover the story because it's just. So. Fucking. Weird.

Backround here: https://pressaudit.org/how-it-works

When LLM stands for "Likely lying, marketing".

Good question. A story is considered Underreported in the US when it meets all of these criteria:

Category = politics, law & justice, society, climate & environment or health Public interest = News is mostly of public interest (Public interest > 0.5) Geographic relevance = US (directly affecting US institutions, policy, the economy, security, or population, or primarily driven by US actors)

The coverage has:

No more than 1 article from a US outlet No more than 3 articles from US outlets and wire services combined At least 2 different non-US regional outlets More than half of all articles are from non-US regional outlets

This is an internal leak The Information reported, as I understand it.

Compare and contrast Sam Altman externally saying the most ridiculous bullshit.

That's kinda the point: The vast majority have no intense use for LLMs that would justify them paying a monthly subscription. But if they don't subscribe, it's not sustainable!

OK well Zitron is citing The Information which is generally solid, but you're kinda burying the lede. According to this, it's 9 million subscribers because...

"OpenAI projects an 80% decline... from 44 million in 2025 to 9 million in 2026" [!!!]

Surely there's reason to be skeptical when OpenAI's own internal projection indicates a massive, massive loss in subscribers?

I'm still inclined to believe OpenAI's projection was too optimistic. The bank report says that average subscriptions only last 7 months, so that would also help explain the rapid churn.

BTW, the number of global WoW subscribers is currently about... 9 million.

Open AI is fudging its wording by saying more than "9 million paying business users rely on ChatGPT for work, and startups, enterprises, and governments" -- IE they're blending together paying consumers with businesses. Also not setting a time frame, so no idea how many of them are recurring users.

Other source is questionable since (for starters) it relies on sampling and mentions a 10% inaccuracy (!) rate.

The 2-3% number is from two banks, based on direct payment card data.

True, but personal use tracks direct consumer adoption, whereas sustained enterprise use is still very much in the air. (See "tokenmaxxing".)

This outcome was totally foreseeable but somehow Zuck and Boz blew billions anyway.

Om was a mentor to me at GigaOM and while writing my first book, teaching me throughout how to understand the tech industry and cut through the bullshit and noise. But more than all that, Om was a very warm, lovely, generous guy.

He knew the business of tech, he was successful at the business of tech, so when he criticized tech, his punches were solid, precisely targeted, and unignorable. The first day I met him, he told me: "I know where all the bodies in Silicon Valley are buried," as he once told me. And he really did.

Tech journalism today is sorely missing fiery thoughtful voices like his.

RE Your predecessor

No one knows who he was, or what he was doing.

But his legacy remains hewn in the HR dock of Stonehenge.

Ah cool. So you replace NPCs with real players as they come in, or add them into the mix?

Kinda cool! What's the max concurrency?

The Internet is not really "divided" BTW, roughly 99% of Hawai'i locals who know about this seem super pissed off. If you're gonna slopify Hawaiian music at the airport, why not just slopify the entire Hawai'i experience?

Yep. It's wild how most reporters just copy/paste these claims without even asking for references.

Like few of them looked at Meta's own stock earnings reports -- which they CAN'T bullshit about, or they might go to prison -- where it was clear that the Quest was hardly selling.

Grateful I got to tell the NYT what I've been ranting about for years:

“They glommed onto the term ‘metaverse’ without really understanding the concept,” Wagner James Au, author of “Making a Metaverse That Matters,” said in an interview. “Their efforts on their metaverse strategy seemed completely indifferent to what previous platforms had learned.”

Gift link BTW: https://www.nytimes.com/2026/03/19/technology/mark-zuckerber...

"The US struck nearly 2,000 targets in the first 100 hours of the operation. That list was not generated on the fly but it was built months in advance. But managing, sequencing, and executing 2,000 targets simultaneously across multiple domains still requires algorithmic support that no human coordination chain can match unaided. More importantly: maintaining a database of 2,000+ target records and keeping every single one correctly classified is exactly the kind of task that gets delegated to automated systems..."

Last week’s new model was a version 2.0 update of Bytedance’s Seedance video generator. As always, one has the feeling that lies are travelling around the world twice before the truth has time to put its shoes on. So here I am for another debunk.

10 favorite products from the 90s and early 2000s, before pop-up ads, subscriptions and privacy issues became ubiquitous in our devices -- and our products still felt like they belonged to us.

"Here are a few theses: After 2.5 years of insane hype, there’s no evidence that current AI is making the design process faster. Good design comes from a broader process, not a one-off conversation – meaning the one-off chat paradigm is unlikely to generate good design for non-designers. AI architecture means it will continue to be worse at designs that are 'out of the training data' – the bold, the novel, designs with very tight constraints."

Update, 3:03pm PT:

Stephenson's Facebook account has been restored within the last 1-2 hours. (I've been checking since the morning.)

But the irony goes on forever.

That Facebook page is the one managed by his publisher.

He also has (had?) a personal one he used that just got suspended.

This is not an anomaly, by the way. I've interviewed many multiple Meta staffers (including senior leaders), and can find little evidence that leadership actually read "Snow Crash" and/or even cared about virtual worlds. Even after spending tens of billions claiming they were building the Metaverse.

Some background: https://nwn.blogs.com/nwn/2025/05/horizon-worlds-meta-horizo...