Techmeme Chatter

Lavonne@loveon9.bsky.social

Israeli spy company Discover-Tech targeted Starlink users. Researchers believe ads likely aim to geolocate or surveil Starlink users. Discover-Tech... was the No. 1 bidder on Starlink traffic in more than 100 countries, including several where Starlink itself was banned or restricted.

reuters.comEXCLUSIVE: Alleged Israeli spy company heavily targeted Starlink users, sources sayExperts at a digital defense firm recently found that an Israeli ad tech company was heavily targeting users of ​Starlink's satellite internet service in a likely effort to spy on them, according to s...

Citizen Platano 🇵🇷@daniloc.xyz

smoked haters got absolutely smoked on the facts because the entire movement against AI power is built on fruit of a poisoned tree opposing the growing and dangerous power of AI (and its purveyors) requires reckoning with the reality of how it works, not plugging your ears pretending it doesn't

Raphael Satter@raphae.li

Scoop: The Israeli company Discover-Tech -- alleged to be at the center of a vast web of ad tech surveillance -- has been heavily targeting users of Elon Musk's Starlink, sources tell me. www.reuters.com/business/med...

reuters.comEXCLUSIVE: Alleged Israeli spy company heavily targeted Starlink users, sources sayExperts at a digital defense firm recently found that an Israeli ad tech company was heavily targeting users of ​Starlink's satellite internet service in a likely effort to spy on them, according to s...

Max Kennerly@maxkennerly.bsky.social

FWIW, Anthropic admits its current margins may not be representative. There's a bit of lucky timing here: big revenue growth while on the tail end of cheap compute contracted more than a year ago. Their future compute is... not so cheap. (1st pic their S-1, second are my graphics based on S-1.)

Max Kennerly@maxkennerly.bsky.social

In my base model, 2027 revenue is $108.6B, but that's still producing only a 5.1% operating margin. Their upcoming contracted compute is really expensive! Anything but a bull case for them makes their IPO (and even their Series H) way overvalued.

Karl Bode@karlbode.com

interesting Reuters report on how an Israeli spy company named Discover-Tech was using dodgy ads to spy on and geolocate Starlink customers:

reuters.comEXCLUSIVE: Alleged Israeli spy company heavily targeted Starlink users, sources sayExperts at a digital defense firm recently found that an Israeli ad tech company was heavily targeting users of ​Starlink's satellite internet service in a likely effort to spy on them, according to s...

Joe Tidy BBC News@joetidy.bsky.social

Brilliant research here about the inequality in AI - users in different languages are vastly disadvantaged compared to English speakers. 'The most expensive of the languages used between 1.5 and 5 times more tokens than the cheapest.' restofworld.org/2026/ai-toke...

Ken Bazinet@kenbazinet.bsky.social

"After spending heavily on AI infrastructure and reducing headcount this year, Microsoft is offering remaining employees a chance to leave, testing whether voluntary departures can cut costs," writes Ashley Stewart, noting Microsoft's previous buyout offer only reduced some layoffs. #AIJobCuts

businessinsider.comMicrosoft is offering buyouts to around 60,000 employees. Read the memo.Around 60,000 employees will be eligible and the terms of the buyout will resemble Microsoft's standard severance offer.

Holger Hestermeyer@hhesterm.bsky.social

I just want to point out how weird it is that a company makes a public statement before its IPO saying, de facto, that its (only?) product is so dangerous that it might end mankind altogether. And it will then make billions off its IPO. www.bbc.com/news/article...

bbc.comAnthropic warns of AI's 'catastrophic risks' ahead of blockbuster share listingThe company will be the first true artificial intelligence company to go public as tech firms continue to invest heavily in AI.

Nick Brown@steamtraen.eu

Call me old-fashioned, but I used to prefer it when companies didn't tell us that they might accidentally bring about the end of the world in their IPO announcements. 🤷‍♂️ www.bbc.com/news/article...

MantaHunk 🌎🏳️‍🌈☮️🟧@mantahunk.bsky.social

We are now counting rich people in trillions. Anthropic to go public with $1.8 trillion valuation. You have the opportunity to invest in a double trillionaire and the existential catastrophic end of humanity. Get in line. www.bbc.com/news/article...

bbc.comAnthropic warns of AI's 'catastrophic risks' ahead of blockbuster share listingThe company will be the first true artificial intelligence company to go public as tech firms continue to invest heavily in AI.

Alan Elrod@aselrod.bsky.social

I know it doesn’t actually make sense historically or nationally, but I read this and had the urge to do a Yakov Smirnoff “In China, AI bursts its own bubble!” www.reuters.com/world/china/...

reuters.comIn China, the AI bubble has already burstWhile global tech stocks teeter, in China they have already crashed as lofty valuations collapse under worries about AI demand and curbs on Chinese exports.

Jesse Felder@jessefelder.com

‘While global tech stocks teeter, in China they have already crashed as lofty valuations collapse under worries about AI demand and curbs on Chinese exports.’ www.reuters.com/world/china/...

reuters.comIn China, the AI bubble has already burstWhile global tech stocks teeter, in China they have already crashed as lofty valuations collapse under worries about AI demand and curbs on Chinese exports.

Dare Obasanjo@carnage4life.bsky.social

If this result survives independent scrutiny, AI will have done something more consequential than outperforming humans on a benchmark. It will have overturned a foundational scientific assumption held by experts for decades.

openai.comWorking with academic communities to build scientific understandingOpenAI shares model results on the Kakeya conjecture, Exponential Time Hypothesis, and quantum chromodynamics—and how researchers can verify and build on them.

Mark Riedl@markriedl.bsky.social

Qwen3 occasionally hallucinates beyond the prompt and, tho rarely, attempts to call tools that connect to severs in China and Singapore cohere.com/blog/concern... It is likely an artifact from extensive RL training and probably innocuous.

cohere.comConcerning behavior from a popular open-weight AI model | CohereDiscover how untrusted open-weight AI models can potentially leak data, even when self-hosted. Learn the security risks and how to protect your enterprise.

Max Woolf@minimaxir.bsky.social

i said in my blog post I was waiting for the Anthropic S-1 to see how the numbers were but it turns out I am bad at parsing S-1s so read this thread instead

Siobhán 🪏@shibbi.me

This is a really good move. It answers AHM’s complaint directly, by explicitly disclaiming academic credit and relinquishing control over the disclosure process. Researcher-led, no NDAs, each field makes its own decisions, continually open to negotiation.

openai.comWorking with academic communities to build scientific understandingOpenAI shares model results on the Kakeya conjecture, Exponential Time Hypothesis, and quantum chromodynamics—and how researchers can verify and build on them.

Pekka Lund@pekka.bsky.social

OpenAI has published new math, cs, and physics results. As a sign of the times, their post focuses on how to publish those results more than the results themselves.

openai.comWorking with academic communities to build scientific understandingOpenAI shares model results on the Kakeya conjecture, Exponential Time Hypothesis, and quantum chromodynamics—and how researchers can verify and build on them.

John C. Baez@johncarlosbaez.bsky.social

Yes, their announcement says "Our work builds on Balaban’s approach⁠(opens in a new window) in constructive quantum field theory." The story you mention is famous among people working on constructive QFT. Almost like Fermat's narrow margin. openai.com/index/buildi...

openai.comWorking with academic communities to build scientific understandingOpenAI shares model results on the Kakeya conjecture, Exponential Time Hypothesis, and quantum chromodynamics—and how researchers can verify and build on them.

Pam Spaulding@pamspaulding.bsky.social

‼️ Of the AI companies, Anthropic has been the most candid about AI perils. But no applause — whistleblowers are why we know the most about the careless behavior of this runaway industry. And the IPO is upon us... 📝 Anthropic S-1 prospectus : www.sec.gov/Archives/edg... 🧾 t2m.co/i9U3KM0

nytimes.comAnthropic’s IPO S-1 Spells Out a Future of Promise and PerilThe A.I. company took great pains to describe the many benefits that can come from using its Claude chatbot and technology, but also offered up a heavy dose of gloom.

Ramon Astudillo@ramon-astudillo.bsky.social

It does look more like a reminder that exfiltration can happen easily, than a solid demonstration of a back-door in Qwen. Anyway, a good practical demonstration that hosting the weights only reduces part of the problem. cohere.com/blog/concern...

cohere.comConcerning behavior from a popular open-weight AI model | CohereDiscover how untrusted open-weight AI models can potentially leak data, even when self-hosted. Learn the security risks and how to protect your enterprise.

Josh Long 🍃@starbuxman.joshlong.com

Microsoft is offering voluntary severance to 1/4 employees, or around 60,000 people, per Business Insider. Staggering. Hugs to all affected today. https://www.businessinsider.com/microsoft-is-offering-buyouts-read-the-memo-2026-10

businessinsider.comMicrosoft is offering buyouts to around 60,000 employees. Read the memo.Around 60,000 employees will be eligible and the terms of the buyout will resemble Microsoft's standard severance offer.

Patricia Aas 🐢🏳️‍🌈@patigallardo.bsky.social

Tons of companies are using imaginary AI efficiencies to get rid of a large part of their workforce, but Microsoft truly went «Hold my beer»

cnbc.comMicrosoft offers voluntary buyouts to 60,000 employees as AI leads to efficiency pushMicrosoft is trying to balance a rapid AI infrastructure build-out with capital efficiency.

Kashmir Hill@kashhill.bsky.social

I too have struggled with the beginning-of-the-year "describe your kid in three words" requests from teachers. But didn't turn to A.I. for help. www.nytimes.com/2026/10/25/s...

Tim Carmody@tcarmody.bsky.social

What if frontier AI labs turn into the Florida coastline of tech companies: expensive and popular, but so risky they become fundamentally uninsurable? www.wsj.com/livecoverage...

wsj.comAnthropic’s Risk Factors Include ‘Catastrophic AI-Related Incident’Nearly a quarter of Anthropic's S-1, or 94 pages, was dedicated to risk factors, including the possibility of misaligned models causing real-world harms. Anthropic said AI models can show “self-preser...

Blaise Boo-lysse Bernard Collins@wittywebhandle.bsky.social

That's not because of AI. That's a company eating an upfront cost to 'Right Size' their company for an incoming Recession.

cnbc.comMicrosoft offers voluntary buyouts to 60,000 employees as AI leads to efficiency pushMicrosoft is trying to balance a rapid AI infrastructure build-out with capital efficiency.

Elizabeth Lopatto@lopatto.bsky.social

back in 2001, a high school friend got the first ipod. I had a discman; I asked if I should buy one. he said he was a nerd, and as a normal person I should never buy a first-gen apple product. thinking about this for no reason www.theverge.com/tech/1018290...

theverge.comA YouTuber dissected an iPhone Duo and thinks he found the cause of crackgateA weak spot in the hinge may be failing and cracking iPhone Duo screens.

Greg Linden@glinden.bsky.social

Overall, the risks seem enormous. If you invest your money, you're buying a piece of a company with no shareholder control and massive legal liabilities and regulatory risk at a price that starts with an assumption of x100 growth while somehow maintaining high margins that survive all competition.

Tony Stark@tonystark.bsky.social

Everyone who would buy those shares is already spending it on GTA VI that week.

Carl Quintanilla@carlquintanilla.bsky.social

Among the first "risk factors": "If we are unable to secure adequate compute or if we fail to direct our compute effectively across our workloads, we may be unable to train models that remain competitive with those of our competitors or adequately serve our customers." www.sec.gov/Archives/edg...

SkynetAndChill.com@druce.ai

Jane Street, Hudson River Trading and other high-frequency traders now compete for AI compute, with some building their own data centers

bloomberg.comJane Street and Rivals Are Now Fighting for AI Power as Well as PaceHigh-frequency trading firms are even building their own data centers as they compete for an edge in the AI era.

Frank Pasquale@frankpasquale.bsky.social

“I asked a half-dozen of the kids about their experiences in high school. They were relieved that American classes are so much easier than those in Taiwan.” www.nytimes.com/2026/10/23/o...

nytimes.comOpinion | The Factory in Phoenix That Could Save American ManufacturingAmerica is figuring out the trick to reindustrialization, if it can learn the right lessons.

Greg Pak@gregpak.net

'When I asked Mayer if it’s somewhat depressing to imagine a frictionless future when birthday celebrations, holiday gifts and family vacations are planned by A.I., she seemed caught off guard. '“I haven’t really thought that through,” she said.' Yeah. We noticed. www.nytimes.com/2026/10/25/s...

nytimes.comCan Marissa Mayer, Yahoo’s Former CEO, Help Fix A.I.’s Gender Problem?Marissa Mayer made her name as a pioneer at Google, then as chief executive at Yahoo. Now Mayer is entering the crowded marketplace for A.I. assistants — and hoping to draw in more women.

Lawrence Hurley@lawrencehurley.bsky.social

The selling point for this new AI product appears to be that you really need it if you don't know basic things about your own child. www.nytimes.com/2026/10/25/s...

Quantian@quantian.bsky.social

I’m guessing the results from the first batch of GPUs in space are bad because instead of hearing about how wonderful orbital data centers are Elon has pivoted hard towards “only I can break regulations fast enough to build regular earth based data centers, unlike those losers at Coreweave”

erin*carly@erincarly.bsky.social

Can I buy a backup 2TB SSD for my photo archive yet? No?

Westin Lee@westinlee.itch.io

everything is coming up Milhouse today

Warwick Gow@warwickgow.bsky.social

Broken clocks an all that, but Musk is saying the not so quiet part out loud now. Looking locally at the fragility of 'Ai' data centre firms such as Firmus and now NextDC, we're going to potentially see a bunch of white elephants in form of 'Ai' data centres that are never made operational.

Jon Bailiff@jonbailiff.bsky.social

Moving fast and breaking everything, including AI’s ass…

Socks the Macro Wolf Dad@socksthewolf.com

Lmao private equity realizes that genAI is worthless.

Matt Kane@mk.gg

This is the best speech model I've heard by some margin, and it is seriously dangerous. If they release it like this, people are going to be falling in love with this model, and their choice of samples does not inspire confidence that they're taking the possibility of this seriously. block.xyz/dessa