Jongens waar hebben we dit aan verdiend. 's Werelds twee marktleiders in discours-creatie (Clav en Eretz Yisrael) vinden elkaar op het snijvlak van geo-mogging en clip-farming. En het heeft gewoon echt alles. De bovenstaande IDF-soldate Shira Braun die zegt een beheerder te zijn van het IDF TikTok-account is, heeft een probleem. Het Israëlische i24News schrijft: "The IDF Spokesperson's Unit said Braun had acted without authorization. "The soldier acted without coordination with her commanders, and her conduct does not meet the standards expected of IDF soldiers," the IDF said in a statement. "The incident is under investigation, and the soldierwill face disciplinary action." En daar maakte AL JAZEERA dan weer een video over jongens wat is de wereld toch mooi soms. Inmiddels is gebleken dat ze uit het IDF social media-team gehaald wordt, en herplaatst is als KOK.
Vervolgens vertelt Clav onderstaand dat hij gedineerd heeft met Netanyahu-adviseurs en dat er besproken werd hoe Netanyahu meer 'gehumaniseerd' kon worden. Waar het antwoord dan op was dat Clav met hem naar de sportschool zou kunnen gaan en dit ge-livestreamed kon worden.
Weer daarna wordt hij stevig aan de tand gevoeld tijdens een Israëlisch tv-interview over die keer dat hij in een club Kanye West's Heil Hitler zong met o.a. Nick Fuentes, de Tate-broers, Sneako en Amrou Fudl, en meer in het algemeen zijn 'relatie' met Nick Fuentes (die vrij miniem is).
En wéér daarna scheldt de dochter van rabbijn Schmuley, degene die bekend staat om haar kosjere sekswinkel, Clav de huid vol tijdens een etentje, ook wegens het zingen van dat Heil Hitler-lied.
Al met al: reden genoeg voor Clav om zijn reis door Israël voortijd af te breken: "People are so negative to you here. It almost makes you want to say, well then I'm not going to come back. That's why I'm leaving". Gelukkig hebben we de beelden nog, allemaal onderstaand.
Echt een gemiste kans voor Israël hoor, Clav kwam er met de beste intenties en dat land kan elke positieve PR gebruiken. Is ze niet gelukt, zonde. Hier nog beelden dat hij verbaal belaagd wordt bij zijn hotel. Een generational fumble, noemen de jongeren dat.
Nederland waterland. Dat ligt niet alleen ontzettend lekker in het gehoor, het ís ook nog eens zo. MAAR VOOR HOELANG NOG? Al tijden wordt gewaarschuwd voor grote watertekorten en zelfs De T. zucht, maar nu brengt Nieuwsuur ons een waarschuwing rechtstreeks van de Landelijke Coördinatiecommissie Waterverdeling, die stelt dat we waterwise zo'n beetje teruggaan naar het jaar 1976 (toen Joop den Uyl nog kraanbeheerder was). We zitten pas een paar dagen in fase 1: dreigend watertekort, maar fase 2: feitelijk watertekort ligt al op de loer. Dat raakt in de eerste plaats scheepvaart en boeren, alleen op een gegeven moment komt ook de burger in de zee van ellende terecht, zeker met die eeuwigdurende hittegolf waar we in zitten. En ja, mogen we dan straks nog wel dweilen? Plantjes water geven? De was doen in een teil? Kan er dan nog muntjesloos gedoucht of moeten we de zweetgeur van Jeroen van sales op kantoor gaan gedogen? Dat worden onsmakelijke toestanden in Nederland. Gelukkig kunnen we in plaats van drinkwater nog altijd aan het bier (toch geldt ook hier: hoelang nog?), verder ziet het er allemaal bepaald niet positief uit.
De hopeloze toestand in het Slavische theater volgt u wellicht in ons onregelmatigheidsklassement van de Weekupdates maar voor deze banger breken we de week graag weer op. Een FAB-3000, een 3.000 kilo zware zweefbom met een explosieve lading van 1.400 kilo die onder een Su-34 jachtbommenwerper geknoopt wordt en dan een kilometer of 40 van het doelwit op deze manier wordt losgelaten om het zelf uit te zoeken. Heel precies zijn ze niet. En dan krijg je dus deze ellende, in Orikhiv, oblast Zaporizhzhia. Op een woonblok, waar weinig mensen meer wonen, dus we zullen het volgens de Code van de Volkskrant wel geen genocide noemen. En er zijn natuurlijk minder Joden bij betrokken!
When it comes to AI services, you don't necessarily get what you pay for. It turns out that AI models with expensive tokens may cost less than models with cheap tokens for particular tasks. And the tooling attached to those models can have a significant effect on cost and output quality. Databricks, which sells data analytics software and services, recently devised an internal coding benchmark to assess the tradeoff between price and performance using various AI models. Matei Zaharia, CTO of Databricks and associate professor of computer science at UC Berkeley, said the company undertook the evaluation because models are often tuned to existing benchmark tests like SWE-Bench – which is "broken," according to OpenAI. Databricks devised its benchmark using real engineering tasks performed by its staff to assess how AI agents perform. Zaharia said while the results reflect the company's internal codebase, other companies should be able to conduct similar evaluations using their own code. One of the things Databricks found was that open weight models like Z.ai's GLM 5.2 are competitive with frontier models, like Anthropic's Opus 4.8. "It landed in the top capability tier, statistically tied with Opus 4.8 on quality, but costing $1.28/task against Opus’s $1.94," the company said in its report. But the price-per-token doesn't tell the whole story. Databricks contends that price-per-task needs to be considered. "Cheaper per-token does not imply cheaper per-task," said Zaharia in a social media post. "For example, Sonnet 5 costs less per token than Opus 4.8 but used more tokens, resulting in higher cost and lower quality." So while Anthropic's Sonnet 5 was around 1.7x cheaper than Opus 4.8 on a per-token basis, it was more costly on a per-task basis – $2.09 for Sonnet 5 compared to $1.94 for Opus 4.8. That's because it completed tasks less often (81 percent compared to 87 percent), and consumed more tokens to achieve the desired result. Academics already reached this conclusion, noting back in March that in about a third of the model comparisons they conducted, the model with the lower listed price ended up costing more. "For example, Gemini 3 Flash's listed price is 80 percent cheaper than GPT-5.4's, yet its actual cost across all tasks is 38 percent higher," they observed. The other thing that had a significant impact on test results was the harness – software like Claude Code, OpenAI Codex, and the Pi coding agent – which passes user input to the model, invokes various tools, and returns results. "Harnesses make a huge difference in cost-performance," said Zaharia. "The very simple Pi harness got the same success rate as harnesses from the LLM vendors with Opus and GPT 5.5, but at 2x less cost!" Zaharia attributed the difference to the size of the input – the context – passed to the model with every turn. When Claude Code served as the harness for Opus 4.8, Databricks measured a context of 742,000 tokens per task, compared to 236,999 for Pi. That's about 3.2x fewer tokens overall. With Codex, the total context per task was 1,235,000 tokens, compared to 665,000 tokens for Pi, which is known for its minimal system prompt. Zaharia said the results explain why Databricks built a tool called Omnigent to harness the harnesses – it's a wrapper for combining and swapping multiple coding agents. It's the front-end equivalent of the kind of back-end model swapping that OpenRouter enables. ®
Seemingly unaware of the concept of irony, Satya Nadella is warning AI-using enterprises to take care not to give away their business secrets alongside the massive piles of cash they’re forking over to frontier labs every month. Writing in a long-form post on X over the weekend, the Microsoft CEO and chairman warned of what he called the “reverse information paradox,” a situation in which purchasers of AI essentially pay for the intelligence product they’re getting twice: once with cash, and again “with something even more valuable,” namely the proprietary business knowledge one has to feed an AI model in order to make it worth using in the rare instance an AI investment actually pays off. “Over time, the information asymmetry becomes increasingly skewed,” Nadella noted. “The seller learns more and more about you as you use what you purchased, while you learn very little about what the seller is learning in return.” The irony is thick, given that Microsoft itself pushes AI that slurps up business data, and Redmond helped get this entire messy AI ball rolling by investing billions into early generative AI leader OpenAI. Azure was the former exclusive cloud home for ChatGPT, and Microsoft leadership arguably helped Altman get his job back when OpenAI ousted him in 2023. The pair’s relationship grew strained in the intervening years, and they loosened several exclusivity provisions in early 2026. It also comes after a number of large organizations paused or restricted Microsoft Copilot deployments in 2024 over a related concern: weak data governance and sprawling internal access rights. Enterprise data security outfit Securiti told The Register in 2024 that about half of the more than 20 chief data officers it polled had grounded Copilot deployments, either switching the assistant off or severely restricting what it could access. The problem was particularly acute in organizations with years of accumulated SharePoint and Microsoft 365 permissions, where overly broad access rights risked exposing sensitive information through Copilot. Fast forward a couple of years, and now Nadella is warning that data protection measures aren’t even enough for a business to stay safe in the AI age. “Models learn from ‘exhaust,’ the prompts people write, the tools agents use, and especially the corrections people make,” Nadella warned. “It's the kind of knowledge a competitor could never buy, and the kind that leaks almost imperceptibly: trace by trace, correction by correction, eval by eval.” Consuming intelligence through AI, Nadella added, creates more organizational intelligence. The Microsoft chief argued that the knowledge generated through those interactions ought to belong to the companies that create it. “Enterprises need a real trust boundary for their human capital and token capital to compound,” Nadella wrote, describing his ideal solution as having “a hard boundary across which nothing crosses, not even the intelligence exhaust, without consent.” In other words, welcome to the post-cloud era when all your AI infrastructure will come home to roost inside your own network. If you think we’re exaggerating, Nadella even mentions that one of the things enterprises need to do to solve the reverse information paradox is to build their own proprietary AI learning environments “within the tenant boundary.” We asked Nadella and Microsoft whether solving the problem goes beyond good data governance, as Nadella suggested in his article, and a spokesperson told us yes, describing the matter as a structural problem with the current generally accepted model of AI business in which companies rely on hosted services. Anyone and everyone using AI for business is at risk, they explained. In addition to isolating learning environments, Nadella’s X note also suggested AI-using businesses should create their own private evaluation systems and retain ownership of organizational AI memory and decouple their orchestration layer from any particular AI model, essentially creating “your own continuous learning loop.” “A company should be able to use a model without giving up the knowledge that makes it unique,” Nadella wrote. The Microsoft spokesperson argued that agent harnesses and memory should be independent of models, and called for enterprises to have the rights to their own usage data and model outputs, echoing Nadella’s comments about the irony of leading AI firms crying foul about model distillation while reserving “the right to learn from customer usage and interaction data.” As for whether this is a generic warning that something in the AI industry’s got to give or a sales pitch with Microsoft positioned as the hero, the spokesperson made that clear, telling us that Copilot and Azure AI Foundry (a hosted solution, it’s worth pointing out) are Redmond’s solution to the problems Nadella outlined in his weekend post. Both separate context, memory, and agent harnesses from AI models themselves, giving businesses an additional layer of assurance that their data is safe, the company told us. It's debatable whether or not Microsoft is actually the AI data protection hero enterprises are looking for. But the bigger point is true: Frontier labs are rolling in valuable proprietary data, and that could come back to bite the businesses that forked it over for free. ®
Is your sex life as private and personal as you think it is? Or is it shaped by – and constantly shaping, in turn – the society and systems you exist in?
This week we’re joined by Dr. Angela Jones, who asks these questions and much more in their new book, Sex in Public. Angela is a professor of Women, Gender, and Sexuality Studies at Stony Brook University. In addition to scholarly works published in many distinguished journals including Porn Studies and The Black Scholar, they’re the author, co-author and/or editor of several books, including Black Lives Matter: A Reference Handbook and Camming: Money, Power, and Pleasure in the Sex Industry which came out in 2020 and advanced and informed a lot of my own understanding of the online adult industry especially.
We get into the history of sexology, how sex toys complicate our understanding of what counts as sex, whether sex needs a definition at all, what happened when they bought and used a sex doll, and why the most vulnerable moments in their book are also the ones everyone wants to discuss.
Become a paid subscriber for early access to these interview episodes and to power our journalism. If you become a paid subscriber, check your inbox for an email from our podcast host Transistor for a link to the subscribers-only version! You can also add that subscribers feed to your podcast app of choice and never miss an episode that way. The email should also contain the subscribers-only unlisted YouTube link for the extended video version too. It will also be in the show notes in your podcast player.
The Los Angeles Police Department (LAPD) announced it will let its surveillance contract with automated license plate reader company (ALPR) Flock expire, becoming the largest police department in the country to drop its contract. Notably, the decision came after an audit of ALPR technology found that, in a two-month period, the LAPD had improperly "investigated" 161 people whose cars were flagged as stolen in the LAPD’s ALPR system but were not actually stolen.
The news that LAPD pulled over 161 innocent people in two months because of improper tagging in the department’s system comes after several high-profile incidents in which people in other states were accosted by police because of data entry or clerical errors in ALPR systems. Joel Feder, an editor of the car journalism website The Drive, detailed a harrowing tale in which he was tracked for days and ultimately pulled over by police in Minnesota because the license plate of the car he was reviewing for the website had been entered into the Flock system as stolen by a police department in California. Monday, the website MotorBiscuit wrote about an innocent woman who was jailed for 13 days because she drove a black Dodge Durango and police searched the Flock system for a Black Dodge Durango suspected of being involved in a fatal hit-and-run accident.
Image: LAPD OIG
A new report by the LAPD Office of the Inspector General (OIG) suggests that instances of people being falsely pulled over because their license plates have shown up on an ALPR “hot list” are very common, and that the surveillance of people on hot lists that ultimately result in no action from police is staggering. Many ALPR systems have this “hot list” feature, which is where police enter a license plate and get a ping or notification about the vehicle’s whereabouts whenever it passes a connected ALPR camera. In a two-month period between August 1 and September 30, 2025, the LAPD’s cameras generated more than 210.5 million license plate reads, according to the report.
“During the review period, officers acknowledged 161 alerts as accurate license plate matches; however, subsequent investigations determined the vehicles were not stolen,” the report reads. “In addition to creating an inconvenience for vehicle owners, these inaccuracies can affect individual liberty interests, erode public trust, and potentially create substantial legal and financial liability concerns.”
The report notes that this happened because of “inaccurate or outdated information, increasing the risk of unnecessary enforcement actions, including vehicle stops and wrongful detentions, or a confrontation with serious consequences,” and that in many cases, license plates remained on a hot list after a stolen vehicle had already been recovered or was reported as not stolen, meaning the cops are in some cases pulling over the lawful owner of the vehicle.
Notably, the report states that when police get an ALPR hot list hit, the department generally considers any subsequent action to be a “high-risk” stop, meaning the risk of confrontation or potential danger is greatly increased from routine traffic stops for running a red light or speeding.
“When a license plate matches with a vehicle of interest on a Hot List, an alert will appear on the police vehicle’s Mobile Digital Computer,” the report reads. “Often, officers will approach the vehicle with extreme caution or conduct a ‘high-risk’ stop. This involves calling for back up, air support, and a supervisor and ordering the suspect out of their vehicle.” The report says, “department policy requires officers to attempt to verify the accuracy of the ALPR alert prior to conducting a stop,” but that often does not happen. The report also states that, on the vast majority of hot list hits, no action is taken by police meaning that specific people are being subjected to tracking and surveillance for no readily discernible reason. In the two-month audit period, 5,911 different license plates were tracked. No action was taken against 4,575 of those cars.
The LAPD said in response to the report that cars improperly flagged as stolen “generally result from the timing of record updates outside of the Department’s control, such as delays by another jurisdiction or a vehicle owner in clearing a plate from a Hot List after a vehicle has been recovered or is no longer wanted.” In other words, LAPD is often relying on other police departments to remove license plates from a hot list, highlighting the problems with networking different surveillance systems together.
The LAPD OIG report, which appears to have directly led the LAPD to allow its Flock contract to expire, studied the use of three different ALPR systems the department has been using, including static, pole-mounted cameras from Motorola and Flock and cameras in police cruisers made by Axon. In total, the department has nearly 2,000 ALPR cameras; LAPD accesses data for both Flock and Axon systems through Flock’s backend thanks to a data sharing partnership between Axon and Flock, according to the report. The report said the department was able to recover 337 stolen cars during the two months and that ALPR data led to 74 arrests total.
Both the OIG and the LAPD determined that the ALPR system needs to be reconsidered. The OIG suggested that the LAPD “suspend the deployment of new ALPR cameras and the execution of new ALPR-related contracts pending public input and a broader reassessment of vendors and data practices” and “strengthen oversight of ALPR data access.” The LAPD allowed its Flock contract to expire over the weekend, and said it would not enter into new contracts until going through a full audit process.
Al sinds zijn verkiezingswinst in april wilde premier Péter Magyar dat president Tamás Sulyok aftrad. Sulyok, die door Magyar werd bestempeld als een „marionet” van de radicaal-rechtse oud-premier Viktor Orbán, weigerde dat.
Microsoft geeft als eerste grote Amerikaanse partij inzicht in het stroomverbruik van zijn datacenters in Nederland. Dat is groot, en groeit de komende jaren bovendien hard. Terwijl andere datacenteruitbaters zwijgen, verhardt de strijd om toekomstige aansluitingen op het stroomnet.
Ongeveer een kwart van de geslaagden op havo en vwo kiest voor een tussenjaar. Ze willen reizen, sporten, een taal leren, vrijwilligerswerk doen. En ze zijn op zoek naar persoonlijke groei. „Ik weet nog niet zo goed wat mijn kwaliteiten zijn.”
Jeongmin Lee is interested in the ways “memory is carried through craft and repetition.” On traditional Korean mulberry paper, or hanji, Lee draws delicate lines in ink and pigments known as bunchae, rendering rippling textures that whirl across the page. Steeped in local folklore and mythology, the Busan-based artist creates surreal scenes that conjure fantastical tales of life by the sea.
“Most of my recent projects begin with reading regional folktales, visiting places connected to those stories, and collecting fragments of history, mythology, and oral traditions,” she says. “I rarely paint a folktale exactly as it’s written; I’m more interested in its symbols, emotions, and the questions it leaves behind.”
Many of Lee’s compositions focus on women’s knowledge, labor, and resilience and how those qualities emerge through coastal storytelling. Tales of powerful sea gods and the diving traditions of the haenyeo commingle into balanced illustrations that translate the symbols and motifs of the region anew.
While her approach is labor-intensive, Lee enjoys utilizing such a meticulous, meditative technique. “Painting with traditional pigments requires a slow, layered process. The colors are built gradually, allowing the paper and pigments to create subtle textures that wouldn’t be possible with other materials,” she adds. “It gives me space to sit with these stories while I paint them.
Many of the works shown here are part of Daughters of the Sea, an ongoing series recently on view at the SĀBRS Festival in Riga. Lee hopes to expand the project into one that’s participatory and connects similar narratives from across the globe. “I’m exploring ways that folklore can become something people experience and talk about together, rather than something that exists only on a gallery wall,” Lee shares.
She’s also currently working on a graphic novel centering on Busan’s mythology and folklore, which she hopes to complete next year. Keep up with her projects on Instagram.
I have a Pi 4 B that, after a power failure, seems unable to talk HDMI in any resolution higher than 1024x768. I have tried both ports, multiple cables, multiple monitors, and a known-good CF card.
# xrandr Screen 0: minimum 320 x 200, current 1024 x 768, maximum 7680 x 7680 HDMI-1 connected primary 1024x768+0+0 (normal left inverted right x axis y axis) 0mm x 0mm 1024x768 60.00* 800x600 60.32 56.25 848x480 60.00 640x480 59.94 HDMI-2 disconnected (normal left inverted right x axis y axis)
I have also tried various permutations of hdmi_force_hotplug, hdmi_group, hdmi_mode, hdmi_force_mode, hdmi_force_hotplug and hdmi_drive in config.txt, and video=HDMI-A-1:1920x1080@60D in cmdline.txt with no change.
Before I toss it in the trash, does anyone know what could have caused this and if it is fixable? Is this a known failure mode of the video hardware scorching itself somehow?
Ik leerde Tiong Ang kennen via Remy Jungerman met wie hij in 2002 een residentie deed in Indonesië. Ik ben zijn werk sindsdien blijven volgen, eerst toen ik meer reisde op afstand en sinds ik [Meer...]
An anonymous reader quotes a report from The New York Times: A state-owned newspaper in China recently published a satellite image of a data center in Gainesville, Va., writing in English that the development of artificial intelligence posed a threat to Americans' physical and financial well-being. A comic strip made to look as if it had been published by a Maryland news outlet -- created with OpenAI's ChatGPT by people in China, the tech company said -- circulated on X this year, blaming data centers for soaring electricity bills. It showed a tycoon smoking a cigar and clutching bags of cash. A video shared on X by a known covert Russian influence operation questioned the viability of a data center that an American company, Firebird, is constructing in Armenia, the small Caucasus nation that has been a focus of Kremlin pressure. "The country's electrical grid instability may render it useless," the video's narrator says.
All are examples of a push by foreign adversaries to seize on what polls have shown is deep ambivalence -- verging at times on hostility -- about the spread of the data centers needed to power A.I. in the United States and elsewhere. China, Russia and, to a lesser extent, Iran have sought to use state media outlets to turn the controversy over data centers in the United States into "a domestic fracture point," according to a new analysis by Alethea, a threat intelligence company, which identified scores of articles and posts on social media this year. These campaigns, whose impact on public opinion remains to be seen, have raised alarms in Washington, where A.I. is seen as a top issue heading into this year's midterm elections.
/s3/static.nrc.nl/wp-content/uploads/2026/07/13193259/130726VER_2035191110_.jpg)
/s3/static.nrc.nl/wp-content/uploads/2026/07/10160057/130726ECO_2035141725_.jpg)
/s3/static.nrc.nl/wp-content/uploads/2026/07/06121134/070726BIN_2034868452_tussenjaar1.jpg)
/s3/static.nrc.nl/wp-content/uploads/2026/07/13170410/130726VER_2035187769_.jpg)
