Google started decades ago with a clever way to focus on trustworthy information among all the crap on the early World Wide Web, but now: "generative AI ... [does] not necessarily know when that source is incorrect." www.theguardian.com/technology/n...

Reminder: All it would take to end the murder of American citizens by an untrained government goon squad is 16 Republicans in Congress voting with Dems to defund ICE (or 23 to impeach and remove Trump — 3 in House & 20 in Senate). That’s it. 23 Americans can vote for the public and end all of this.

Hm Gmail filters seem suddenly completely busted

Jeff Bezos, who is currently worth $250 billion, spent $500 million on a yacht. He could give the Post all the resources it needs until the end of time. If you're going to kill one of America's great newspapers, why not just sell it to someone who gives a crap?

Them, but also more directly: 285 people out of 535 could end all this (218 + 67). That fact should not be treated as an "oh they'll never do it" shrug, as if they don't have agency. They should be hammered constantly as traitors and cowards every day they don't.

If banks can't easily sell off the debt they are putting into data centers they are going to stop putting debt into data centers, it really is that simple. This story tries to soften it by saying there's "still investor interest" but then the actual facts say otherwise.

From Business Insider - Oracle's $38 billion loan closed, but those who have bought into the various loans around OpenAI's data centers can't sell off the debt without taking a loss - and the pool of investors who still have an appetite for their debt has shrunk. archive.is/WfUfe#select...

hey did you see a headline on google? turns out that headline might be an llm hallucination www.theverge.com/tech/865168/...

It’s not good, folks. It is, in fact, very bad.

Do not store your Bitlocker encryption keys on Microsoft's servers if your threat model includes governments or law enforcement. As this article points out, this is the result of a design choice Microsoft made. It didn't have to be this way. www.forbes.com/sites/thomas...

You can look at the ways that LLMs fail -- how for example they say the character held hostage in the motel is a motel clerk, not a bank clerk, because it overreads the setting as linguistic context, etc.

Again, the only thing that will be more embarrassing than being the first institution to self-censor will be being the last one. Good job everyone.

Trump tearing down the East Wing and then being unable to build anything in its place would certainly seem like potent symbolism www.cnn.com/2026/01/22/p...

The new administration has knee-capped the formerly rapidly growing U.S. clean energy manufacturing buildout, handing this sector to China and other countries. A big loss for our economy. www.canarymedia.com/articles/cle...

Telling people who fought and died by your side, after you got attacked and asked for help, that they're ungrateful moochers who never do anything for you, is really low. False, rude, and self-harming all in one. Denmark and other NATO allies are right to be appalled. I'm not sure what else to say.

In dozens of interviews, sociologist, UX researcher, and interaction designer Diana Enriquez finds middle managers forced to "feign AI success" in a weak job market—for instance, pretending error-free drafts came from AI tools that actually failed, creating burnout rather than productivity gains.

remember: the rules of "CEO said a thing!" journalism require you offer no historical context about the CEOs past statements or actions, you downplay or ignore any realistic challenges to the CEOs claims, and, most importantly, you avoid citing any actual experts in the field you're writing about

this is honestly a very good metaphor

My new book — Clearing the Air — comes out in the US on February 19th 📚 Goodreads is running a giveaway, in case you want to enter and have a chance of getting a free copy! @mitpress.bsky.social www.goodreads.com/giveaway/sho...

This is an elected member of Congress & the GOP saying that he welcomes a war against another nuclear power who is literally our oldest ally & with whom we have a formal treaty military alliance. Congrats again on that National Conversation™️ about which party has moved further from the center 🤷‍♂️🤦‍♂️

In dozens of interviews, sociologist, UX researcher, and interaction designer Diana Enriquez finds middle managers forced to "feign AI success" in a weak job market—for instance, pretending error-free drafts came from AI tools that actually failed, creating burnout rather than productivity gains.

Good article, but I think it's wrong to suggest there aren't already hordes of social media accounts controlled by propagandists feigning consensus and shouting down the opposition. It's already overwhelming now. You shouldn't trust that most random accounts commenting are who they claim.

Political leaders could soon launch swarms of human-imitating #AI agents to reshape public opinion in a way that threatens to undermine #democracy, a high profile group of experts in AI and online misinformation has warned. (Guardian)

Social media platforms are not “digital town squares.” They are adversarial spaces, actively manipulated and distorted by bad actors intent on dividing the country and undermining our national security interests.

Not great Bob.

Allen School researchers earn NeurIPS Best Paper Award for revealing the ‘Artificial Hivemind’ effect across LLM open-ended generation news.cs.washington.edu/2026/01/22/a...

“.. Here’s the reality: Tesla didn’t remove the safety monitor. It simply moved them from inside the Robotaxi to a chase car following closely behind.” 🤡 @electrek.co #$TSLA x.com/electrekco/s...

how many more times does Tesla need to make it clear that it's putting on performances rather than seriously developing real driverless technology before Wall Street pulls its head out of its ass?

The mRNA vaccines, customized against high-risk skin cancers, appeared to reduce the risk of cancer recurrence and death by nearly 50 percent over five years when compared with standard treatment alone.

The curl project plans to end its HackerOne bug bounty program at the end of January, citing a surge in low-quality AI-generated vulnerability reports (Lawrence Abrams/BleepingComputer) Main Link | Techmeme Permalink

Today I am publishing what I believe to be the most detailed available map of Trump's job approval rating. Powered by over 12,000 interviews for our Strength In Numbers/Verasight poll. I hope this generates many stories, and hours wasted exploring data! www.gelliottmorris.com/p/a-very-det...

Three AI executives admitting everyone hates their product this week.

I would also note: X has deleted a handful of these images. But the vast majority of them remain live on X today.

"Applications are now open for the Disinformation Summer Institute. This 4-day event (June 15-18) is intended primarily for advanced PhD candidates, advanced law and medical students, post-docs, assistant professors & early career researchers seeking to better understand and address disinformation."

HCDE Professor Kate Starbird has been named an ACM Fellow, one of the highest honors in computing. She's recognized for contributions to understanding and improving information ecosystems, including during crisis events and addressing misinformation. www.hcde.washington.edu/news/article...

Congrats to @cip.uw.edu co-founder @katestarbird.bsky.social!

Stand up to bully, give a firm no, threaten retaliation and take steps to make it credible — he backs down, and tries to save face by claiming pre-demand status quo was his idea. Appease bully, treat his demands as reasonable, appeal to mutual interest — he sees you as an easy mark, demanding more.

The new NYT poll finds Trump 17 points underwater on immigration, and the nonwhite and young voters who went to Trump in 2024 have snapped back to Dems It's almost as if 2024 might not have represented a seismic cultural realignment on immigration after all, despite 1,000 NYT op eds to the contrary

Repeat after me - the institutions of this economy depend on the institutions of this democracy, among which is the rule of law

I know I am a tediously broken record on this, but Smith's point (& mine) cannot be repeated often enough. I/we will stop when the couple of dozen Republicans in Congress who can end this all now finally stand up & do their damn jobs. No, we can't wait until 2028. It needs to happen now.

Jack Smith: "My fear is that we have seen the rule of law function in our country for so long that many of us have come to take it for granted. The rule of law is not self-executing. It depends on our collective commitment to apply it."

It is amazing how far Google is going to make their products worse. They had total market domination and now I am actively looking for ways to stop using it.

“One notable MIT study found that 95 percent of companies that integrated AI saw zero meaningful growth in revenue. For coding tasks, one of AI’s most widely hyped applications, another study showed that programmers who used AI coding tools actually became slower at their jobs.”

"Putting data centers in space pushed the billionaire toward an IPO, sources say" is a hell of a way to say "he's trying to squeak in an IPO before the window slams shut, as the AI hype cycle reaches an absurdly unsustainable denouement"

The disconnect between management and employee on AI usage is big. At the same time, CEOs are reporting they aren't seeing any revenue growth from AI. 🤷 www.wsj.com/lifestyle/wo...

If you're only reading headlines and see "Trump Says He Won't Use Force" it will give you a misimpression of both how bellicose this speech is and how dug in he is on Greenland.

Science personnel and funding gutted at astonishing scale in the US: www.nature.com/immersive/d4...

Your regular reminder that Republicans in Congress approve of what the President is doing.