Boy, do I feel silly. Until now, I've been a passionate supporter of SS-Sturmbannführer Arnold Toht, the Nazi who got his face melted off by the sight of God, in his race for Senator. I believe the people of Maine deserve better than Susan Collins, and so even though I had some misgivings about Toht, I ignored every click and whirr of my moral compass to cheer him on.
Previously, previously, previously, previously, previously, previously, previously.
Grok, what's the difference between right and wrong?
Last Sunday, San Francisco District 6 Supervisor Matt Dorsey relaxed after what was, for most, a fun-packed SF Pride by repeatedly prompting xAI chatbot Grok to provide justifications for the killings of civilians in Gaza by Israeli forces. [...]
In lieu of actually addressing his critics online, of which there were many in the replies, Dorsey only had eyes for Grok, a chatbot that is notorious for being programmed to push false narratives about politics, history, and everything else according to the very transparent whims of xAI's trillionaire backer, Elon Musk.
Matt, are you OK? I really mean that. I'm not trying to be funny. Frankly, I tried to write something satirical about what was happening -- tried to find a lighter way through this -- but the more I dug in, the more concerned I started to feel.
Only a year ago, some backend overcompensation by Grok's puppeteers resulted in the bot compulsively spitting out conspiracies about "white genocide" in South Africa, praising Adolf Hitler, and engaging in Holocaust denial. To that point, AI researchers have warned that AI boosters' framing of chatbots as objective and reliable could be weaponized to push propaganda and even encourage people to commit violence.
So of course, watching an elected official openly using it to confirm his very loaded questions about the correctness of killing civilians is unsettling -- even terrifying.
In his most recent dealings with Grok, the supervisor's behavior is giving off the sketch vibes of a person antagonizing those around him while not even looking at them. Instead, it's almost like he's staring off into space and talking to someone who isn't there.
As it stands today, X is a platform that actively facilitates the turning of brains into pudding. And if this were some rando in the city posting, it would be easy to dismiss them as another soul lost to AI obsession. But Dorsey represents a major and heavily populated slice of San Francisco, including Mid-Market, Mission Bay, SoMa, and other neighborhoods. Supervisors like him write legislation and have a profound impact on the lives of the people in their districts and the city at large.
If this is what he's doing publicly, what is he asking Grok in private?
Despite this, Matt remains only San Francisco's second-worst weird dude named Dorsey.
Waymos were a key reason for the traffic snarl in the Presidio after Saturday night's foggy fireworks display on the Golden Gate Bridge. Videos and social media posts from people stuck in traffic show Waymos locked in giant lines, randomly blocking the road, and stuck in a roundabout.
A number of Waymos had to be towed after they lost battery power; the vehicles that could hold a charge made errors on the road that backed traffic up for blocks, bystanders said.
Fellow drivers acted like, well, drivers. "We realized people were getting out of their cars, yelling and screaming at these Waymos because there were no drivers," Dave Guingona, who was stuck in Presidio traffic for two hours, told NBC News. [...]
In the Mission, meanwhile, one Waymo drove into an intersection despite a large lit firework sitting in the middle of the road. Video taken inside the car by passenger Rose Peterson shows the firework's sparks flashing all around them as the Waymo takes a lazy left turn. A second angle shows Peterson's Waymo tip the firework over, which then shoots projectiles at cars, homes, and bystanders on the sidewalk. [...]
A second Waymo that encountered a firework was a different story, though: A vehicle that drove into a firework near the 1200 block of Connecticut Street in Potrero caught on fire, Waymo said. The Waymo was unoccupied and there were no injuries; the vehicle had to be removed by the San Francisco Fire Department and other city authorities, according to Bonelli.
These incidents raise serious concerns about autonomous vehicles, especially those operating during unusual traffic patterns that may require real decision-making by drivers. "Human drivers don't just drive. They are the captain of the ship and anything that can happen to it," AV expert Phil Koopman told Gazetteer in 2024. "Well, if you put a computer in charge and it's not capable of doing anything other than navigating traffic, who's the real captain of the ship?"
The office of San Francisco District Attorney Brooke Jenkins did not respond by press time to Gazetteer's questions about legal mechanisms that might hold Waymo accountable when its operational failures, including parking and moving violations, lead to blocked roads and other forms of extended congestion. Jenkins recently attempted to prosecute seven members of the so-called "Golden Gate 26" who blocked the bridge in 2024 during a protest and were charged with felonies for false imprisonment (for keeping people stuck in traffic) and obstructing a thoroughfare.
Meanwhile, Mayor Daniel Lurie, who has been a major ally for Waymo in its San Francisco expansion, did not mention the company in his statement regarding the July 4 pileups.
Mirroring these videos that were posted on oligarch-controlled social media sites for posterity...
I have so many posts about Waymos on fire that it almost deserves its own tag:
Previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously.
Bernie Sanders is campaigning to delete the part of the military spending bill that would further integrate the US and Israeli armies.
The pretend intelligence ChatGPT Health is presented as a good source of medical advice, but an independent study submitting test cases found it often failed to recognize dangerous emergencies. It responded, "Make an appointment in the usual way," when (based on the described case) it should have responded, "Go to the emergency room now!"
This was not due to mere ignorance. The system's behavior showed it was influenced by irrelevant details included in some test cases as a red herring.
In the dispute between Rep. Dan Goldman and the Poetica Cafe, there is wrong on every side.
Judging from Goldman's Wikipedia page, it is clear that he excuses Israel's atrocities in Gaza, which may have killed hundreds of thousands of Palestinians and seem to have targeted journalists and medical doctors, citing HAMAS's equally cruel but far smaller atrocities.
The store was within its rights to refuse to serve him for his political views, which he has stated publicly. Both Goldman, and magat official Harmeet Dhillon, are acting unjustly by equating politically based refusal with antisemitism. The systematic persecution of people who criticize Israel and the labeling of them as "antisemitic" is a grave threat to freedom of speech in the US.
The Guardian article mentioned that Goldman attended a rally for Israel which Smotrich also attended. It then proceeds to blame Goldman for Smotrich's crimes and views — but the article presents no justification for that leap. Whether the rally endorsed Smotrich's practice of driving West Bank Palestinians out of their homes and lands is not stated in the article.
Hegseth directed a military training base into a flu outbreak by revoking the rule requiring vaccination.
Requiring recruits to follow medical safety precautions is generally wise, because it protects them all from illness. But requiring vaccination has a second potential benefit: it can keep some credulous and dangerous right-wing extremists out of the military.
*[Continuing steady] ocean heating fuels "staggering" loss of marine life [over time], study finds.*
A research study found a link between prenatal exposure to PFAS with the likelihood of later developing a medical problem in the ovaries.
Steadily growing repression is settling over Georgia under a pro-Russian promoted by Putin.
He was elected, and then showed his true colors.
The UK has adopted a policy of regulating specific giant tech platforms.
The requirements on Google regulations described there seem like a step forward to me, because they can reduce Google's power over the public.
However, don't fall into the trap of thinking of this issue on the basis of copyright (misleadingly referred to in the article by the misleading term "intellectual property" — see https://gnu.org/philosophy/not-ipr.html/). Copyright is not a principle, it is an ad hoc solution for a situation in the past that has been adapted ad hoc for subsequent changes. (See https://gnu.org/philosophy/copyright-vs-community.html.)
Please join me in rejecting the term "content". It treats publications as fungible, a sort of commodity, and that devalues each and every published work.
*The UK’s social media ban for under-16s has just empowered big tech [even more].
Age verification means that the sector’s biggest players will now have access to information that will only make them richer and more powerful.*
A mere "rule change" could put US-funded scientific research under the detailed, specific control of ignorant political officers, functioning as a whole regiment of Lysenkos.
And you can get a discounted seated "Season Pass" ticket that gets you a seat at all five events through the end of the year!
As just a few examples of "great movie, but bad fit": Aliens, Strange Days, The Thing. They're great movies! But they aren't party movies. They're too good.
On the other end of the scale is the obscurity problem: I'd love to do Barbarella, Danger Diabolik, Dr. Phibes, but nobody's heard of them. The 6 people who would show up (8, counting me and Kingfish) would love it, but it would be a flop.
Anyway, we've got a list, and if these next five movies go well and we start getting some regulars, maybe we'll be able to dip into the backstock as well. So if you want us to keep doing this kind of thing... vote by showing up.
Happy July 4th! For those of us around the world contemplating independence, it's a good day to think about how we came to rely on expensive cloud infrastructure for our fundamental computing needs.
With that in mind, here is my latest toy project: an open source tool that makes replicating, forking, sharing, and running container snapshots fast and easy across cloud and personal devices.
It's fun to play with, especially on bare metal hardware you run at home, or rent from a provider like Hetzner or OVH. Or, because it uses Tailscale, why not all of them in a single mesh?
There's a lot more to say but I don't have time right now. Details are in the README.
I will say this: humans and AI agents both want the same things when they're trying to get work done. Ephemeral containers aren't really it. But how about unlimited disk space, fast CPUs, an undo button, and the ability to move to whatever provider offers the best hardware at the best price? That's more like it.
Go visit thundersnap on github and tell me what you think!
Please be aware that I have the most comprehensive collection of jokes of this formation, so you cannot stump me, that's not what I'm asking. I have 100% heard your joke before. I just want to hear your best formation. Bring it.
Previously, previously, previously, previously, previously, previously.
The 40th Annual Conference on Neural Information Processing Systems (NeurIPS) -- which is slated to take place in Sydney, Australia, in December 2026 -- bans peer reviewers from uploading papers they referee to AI chatbots, as the practice breaches confidentiality. [...] To enforce the policy and catch illicit AI use in peer review, the event's organizers have included deliberately concealed instructions for large language models (LLMs) in papers sent out for peer review.
The instructions tell an LLM to use telltale phrases -- such as "This work addresses the central challenge" and "The claims of the paper" -- in a peer-review report. [...]
"Designing a trap that presumes bad faith corrodes the relationship the whole system depends on," Sören Auer, a computer scientist at Leibniz University Hannover, wrote on LinkedIn. "You do not build a healthy reviewing culture by treating your reviewers as suspects."
But others see merits in the approach. A similar prompt-injection effort has caught hundreds of reviewers misusing LLMs in submissions for next week's 43rd International Conference on Machine Learning (ICML 2026) in Seoul, South Korea, according to Nihar Shah, a computer scientist at Carnegie Mellon University and scientific integrity chair of that conference.
Culture of trust, uh huh.
In 1929, Secretary of State Henry Stimson closed down the "black chamber" -- the State Department's code‐breaking office -- on the principle that the way to make men and nations trustworthy was to trust them. As he later told aide McGeorge Bundy, "Gentlemen do not read each other's mail."
Planet Debian upstream is hosted by Branchable.