Feed aggregator

June emerged from stealth today with a $20 million pre-seed round to make AI adoption simpler.

This story was originally published by CalMatters. Sign up for its newsletters.

In the past year alone, they’ve helped catch suspected murderers, bank robbers and hit-and-run drivers. They’ve located dozens of stolen cars and helped send a wildland arsonist to prison. 

They’re also distrusted and despised by privacy rights activists on the left and on the right. Just last month, police accused a suspect in Monterey County of yanking down three of them with his pickup.

They’re called automated license plate readers. Every day, untold numbers of Californians drive past the thousands of cameras that at least 230 police and sheriffs’ departments have posted along roadways or installed on their patrol cars. The cameras in real time capture and read plate numbers, then cross-reference them with law enforcement databases of vehicles police want to locate, such as stolen cars, getaway vehicles or those that may be carrying a missing or wanted person.

Since 2019, California Democratic lawmakers have unsuccessfully sought to further restrict how police use the technology. They’re trying again this year with a bill that would add limits on how long police may retain license plate data, how they access it and with whom they share it. 

Police agencies and officers’ unions across the state oppose the latest measure, which is similar to a bill Gov. Gavin Newsom vetoed last year.

“This bill is about one simple principle: Protecting the privacy and safety of Californians while ensuring law enforcement tools are used responsibly before it’s too late,” the author of this year’s legislation, Riverside Democratic Sen. Sabrina Cervantes, told the Assembly privacy committee last month. She did not respond to a request for an interview from CalMatters.

Critics of the technology cite high-profile examples of California police using license plate data inappropriately, including to stalk or harass innocent people. As CalMatters reported last year, California officers have repeatedly violated an 11-year-old state law that prohibits sharing the data with federal authorities and others outside the state. 

404 Media also reported last year that a Texas cop searched a national license plate database to try to locate a woman who had an abortion, leading to fears that conservative states that criminalize abortion are accessing California’s plate data to hunt down patients or providers.

Cervantes’ bill would generally limit to 30 days the amount of time agencies could keep license plate data, unless it pertains to an active investigation or is part of a search for missing people. It also sets limits around who can access the data and adds training and search-tracking requirements. 

Police and sheriffs’ departments would be prohibited from entering into contracts with camera companies that provide default access to plate databases that federal and out-of-state police can search.

That provision is in response to Flock Safety, a prominent license plate reader company that allowed federal authorities to search local license plate data as part of a pilot program, said Tracy Rosenberg of Oakland Privacy, one of the bill’s supporters.

“[Police departments] learned after the fact, like our advocates did,” Rosenberg said. “And we all said, ‘What the …’ You know, pick your four-letter word.” 

Flock Safety did not respond to CalMatters’ request for an interview, but the company says on its website that it doesn’t work with immigration agencies and it stopped its federal pilot programs. Local agencies have total control over the data Flock readers collect, the company said.

“By default, sharing with federal agencies is disabled,” the company wrote. “ICE does not have direct access to Flock cameras, systems or data.”

Why police are opposed

At least 39 police agencies and officers’ unions oppose Senate Bill 1013. Combined, they have spent at least $5.6 million on legislative campaigns since 2000, according to the CalMatters Digital Democracy database.

Police say the technology has proven invaluable at helping them solve crimes.

Tustin Police Chief Sean Thuilliez, president of the California Police Chiefs Association, said the technology helped clear an innocent man. It also identified a suspected killer when a woman was shot and killed on a street in the Orange County city in May.

The woman had just testified against her ex-husband in a restraining order case, making him a prime suspect, Thuilliez said.

But Thuilliez said that thanks to the city’s 102 license plate readers, eight of which are from Flock, detectives quickly ruled out her ex-husband and were able to apprehend her former boyfriend at the Mexico border. He’s now charged with first-degree murder.

“It’s technology mixed with good investigative work,” he said in an interview. “But within one weekend, we were able to arrest the suspect for this person’s murder.”

The Berkeley Police Department noted similar successes after 52 Flock cameras were installed last year. In a report to the Berkeley City Council, the department noted license plate readers had been used in 121 cases, contributed to 58 arrests and led to the recovery of 37 stolen vehicles.

The department noted its rate for solving robbery cases climbed from 34% in 2024 to 49% the following year, thanks in part to the license plate data.

“The technology has fundamentally improved how officers patrol and how detectives investigate crimes spanning a range of serious offenses including robbery, kidnapping, carjacking, burglary, weapons violations, organized retail theft, stolen property and DUI,” the report said.

Why Newsom vetoed last year’s bill

Such investigative success stories may have helped persuade Newsom to veto last year’s bill, which would have required police to dump their license plate data within 60 days.

In his veto message, he said the restrictions on sharing data and forcing departments to purge it could hinder officers’ ability to solve cold cases and find missing people.

State Sen. Sabrina Cervantes at the Capitol in Sacramento on Jan. 5, 2026. Credit: Miguel Gutierrez Jr. / CalMatters/Pool

Newsom also said there was no money in the budget to pay for the mandatory California Department of Justice audits last year’s bill would have required.

To address Newsom’s concerns, this year’s proposal allows police to retain data to locate missing people and it makes a state audit program contingent on future funding.

Newsom’s press office declined to say whether he has a position on the latest bill, which has advanced through the Legislature with only Republicans voting against it. The measure is now pending before the Assembly Appropriations Committee.

Meanwhile, at least one Republican lawmaker is taking the issue to Congress. U.S. Rep. Thomas Massie of Kentucky recently shared a meme of a “Don’t Tread on Me” snake wrapped around a camera. The riff on the yellow Gadsden flag says, “Don’t Flock me.”

“I’ll soon be sponsoring a bill to withhold federal money from municipalities and police departments that deploy Flock (and other style) cameras to surveil law-abiding citizens,” Massie wrote on Facebook. 

Rosenberg of Oakland Privacy said she understands the cameras’ usefulness to California police, but she thinks they’ve become ubiquitous. Police, she said, also shouldn’t have a permanent pass to keep “mountains” of data for as long as they want. 

“There is kind of a balancing act. We as human beings also need some camera-free space,” she said. “We don’t want to be on camera every minute of every single day.”

Have something to say? Lookout welcomes letters to the editor, within our policies, from readers. Guidelines here.

The post ‘Don’t Flock me’: California, U.S. lawmakers debate license plate reader technology appeared first on Lookout Santa Cruz.

Oracle has told contributors to OpenJDK – the open source Java project it stewards – not to submit code or other material generated by AI. In a post tagged "legal," Oracle said it was drafting a full policy on generative AI but had set interim ground rules. It cited risks involving review workloads, safety, security, and intellectual property. "Contributions in the OpenJDK Community must not include content generated, in part or in full, by large language models, diffusion models, or similar deep-learning systems. Content, in this context, includes but is not limited to source code, text, and images in OpenJDK Git repositories, GitHub pull requests, e-mail messages, wiki pages, and Java Bug System issues," the post said. Big Red said developers could still use LLM-based tools privately to "help comprehend, debug, and review OpenJDK code and other content, and to do research related to OpenJDK Projects," provided they did not contribute material generated by those tools. The post elaborated on the potential burden for reviewers and the safety, security, and intellectual property concerns surrounding AI-generated contributions. Oracle noted that OpenJDK is the primary Java implementation for many organizations and underpins mission-critical systems worldwide. "Safety and security are paramount. Plausible-looking but incorrect code would put these critical properties at risk," the vendor said. The caution contrasts with Oracle's enthusiasm for AI-generated code inside its own walls. "The code that Oracle is writing, Oracle isn't writing," co-founder and CTO Larry Ellison told Oracle AI World 2025. "Our AI models are writing. We just tell the model what we want the program to do, and then the AI comes up with a step-by-step process to actually do it. We don't write the procedure. We declare our intent, but the model writes the step-by-step procedure; that thing we commonly think of as a computer program." Earlier this year, co-CEO Mike Sicilia said: "AI tools and their coding capabilities would be a threat if we weren't adopting them, but we are and very rapidly. The use of AI coding tools inside Oracle is enabling smaller engineering teams to deliver more complete solutions to our customers more quickly." AI also featured in Oracle's explanation for cutting 21,000 jobs in June. "Deployment of AI technologies across our operations have resulted, and may continue to result, in reductions to our workforce," it said in a statement. Oracle also claims that agents embedded in its applications can reason, make business decisions, and act on behalf of users. Why AI-generated code is suitable for Oracle's products but not for OpenJDK contributions is therefore puzzling to The Register. Presumably Oracle has found some cast-iron method of making its internal AI-generated code safe and secure. It has yet to share the secret with us, but we have asked. Publicizing doubts about AI's ability to act reliably on our behalf could be awkward for Oracle right now. The database group said last month that it would invest $70 billion in the coming year to finance its datacenter build-out, up from $55.7 billion in fiscal 2026, which ended in May. Rating agency S&P responded by downgrading its credit rating to BBB-, just one notch above junk status, saying there was an uncertain path to making money from the investments. Oracle is borrowing heavily to fund its building program, while accepting negative cash flow. More recently, the spread on Oracle's credit default swaps has spiked, indicating that insuring against a default has become more expensive. ®

Donald Trump unleashed an epic tirade at Senator John Cornyn for blocking his IRS lawsuit settlement slush fund, which would let Trump reward allies and grants the Trumps a form of tax immunity. Cornyn is holding up the nomination of Todd Blanche as attorney general until it’s killed. Trump gloated that he’d endorsed Cornyn’s victorious primary opponent and seethed that the American people want him to have this corrupt fund. In response, another foe, Senator Thom Tillis, landed a hard blow, asserting that Trump’s rage over the fund is exactly why it must be killed. Humiliatingly, Tillis treated Trump like a toddler who must be told, “no.” Plus, a defeat would show that Trump’s license to corrupt DOJ is not limitless. (Underscoring this humiliation, after we recorded, Trump petulantly exploded again, saying he absolutely will have his fund.) We talked to New Republic staff writer Emma Janssen, author of our great new politics newsletter all about Democrats, The TNR Blue Book, which you should subscribe to, folks! We discuss her installment on how Dems are exploiting these splits, the deeper reasons for them, and how they reveal Trump’s diminishing power more broadly. Listen to this episode here.

OPINION The UK's National Museum of Computing at Bletchley Park (TNMoC) is a natural center of digital archaeology. It has restored, recreated, and documented landmark machines, from the wartime Colossus computer through to British mainframes of the '60s and '70s. Its collection from the dawn of the microprocessor age also attracts those who live to investigate such things. LLMs, still in their metaphorical diapers, might not sound like a good fit for the hallowed huts. But boy, are they ever. TNMoC has a YouTube channel, and puts out guest talks from anyone with something interesting to say about industry history. These include subspecies of retro geek who reverse-engineer long-abandoned systems, chips, and backwaters of digital history. They thoroughly deserve the sobriquet of digital archaeologists, as they display the profession's expected love of the obscure and the skills required to understand it properly. Their talks bear comparison with reverse-engineering presentations at DEF CON and Black Hat, peeling back layer after layer of mystery to reveal something unlikely ever to affect most of us. Of late, the TNMoC digital archaeology talks have rather suddenly sprouted a new theme – LLMs demonstrating surprising prowess in the art and science of rebuilding the past. One such was bringing Inmos Transputer systems back to life by designing an accelerator for the Acorn Archimedes, the first Arm-based computer. Retro's like that. Transputers were a 1980s UK effort to commercialize low-cost parallel processing. Programmed using their own brain-mangling specialist language, occam, Transputers were unparalleled if you understood how they worked and had a parallelizable task you could occam. Almost nobody did, and the technology was swept away by the time of the Pentium. There isn't much information available about the systems. Phil Pemberton, who designed the accelerator and gave the talk, admitted that he struggled to get his head around occam and doubted many users could have managed without high-level support from Inmos. But Claude Code could, getting his software working. An even more impressive Claude-powered reverse engineering effort, if marginally less obscure, was the disassembly and annotation of Acorn's 6502-based Network Filing System, NFS. This required a deep understanding not only of the code, but of Econet's LAN hardware and software – developed around 1980, before Ethernet became commonplace on microcomputers – plus much of the host hardware and software. The motivation was to write an Econet game, of which very few exist. Creator Rob Smallshire described the undertaking as a project of "infinite scope and no budget." Retro's like that – and it's not alone. The result is an impeccably structured and commented series of source code files, much better than ever existed before. The full story of how that came about is complex and absorbing, but boils down to Claude Code running and debugging emulators, spotting mismatches with chip datasheets, interpreting circuit diagrams and handbooks, simulating machine states, absorbing control flow graphs, and more. This resembles the way an expert human reverse engineer works and, in the right hands, Claude Code appears to deliver a dramatic increase in efficiency without sacrificing quality. A ROM can be processed in a day, fully commented, even to the point of Claude Code saying "Oh, that's clever" of particularly notable bits. "As a big AI skeptic, it just blew my mind," said a voice from the audience. TNMoC audiences are hardcore. This is not simply retro geekery; it is true digital archaeology with true modern implications. There's plenty of precedent. Forensic science owes a great deal to archaeology, such as excavation methodology, taphonomy – the study of what happens to organic remains after death – osteology, and the analysis of mass graves and conflict sites. Paleoclimatology informs modern climate science. And if you want near-infinite examples of how history has valuable lessons for every aspect of the modern day, just ask a historian. Best get comfortable first, though. There is an exact parallel here between digital archaeology and the forensic analysis of the contemporary crime scenes found in so many enterprises. It seems cheaper at the time to patch and wrap old systems to work in new environments than to go through the cost, disruption, and risk of rebuilding from scratch. People move on and documentation, such as there is, is lost, even as projects stay in place. Anyone with a few years in the engine rooms of corporate IT will know the right things to do to responsibly evolve and the many reasons why the wrong thing happens anyway. When migration, merger, reorg, or a new C-suite broom arrives, the skeletons come dancing out of the servers. Now, it seems, there are new ways to keep dem dry bones under control. A technology that can put flesh on the fossils of systems that vanished in the days of dial-up should have plenty to say about more recent antiques, for which far more code and documentation survives. As the digital archaeologists have shown, it's not a zero-effort magic wand. "Here's our server map, document it and tell me what to do" is an incantation for slop. Work with it with logic, reason, and iteration, though, and you'll be able to build the legacy system encyclopedia everyone needs and nobody has. Given enough context, LLMs are very good indeed at filling gaps, to the extent that critical, skeptical experts are regularly surprised and gratified by the results. They know what to do and what a right answer looks like, and thus can provide the context for the model to do the hard work. In the case of badly documented, poorly understood, and generally mysterious digital technology, we have the proof of concept that LLMs cut the mustard. You still need the skills and experience to make this fly. Don't wait for a deadline or a crisis: crack open a CLI and start experimenting on the ugliest monster in your life. Take care over security – mixing LLMs with corporate secrets is a live-fire exercise – and make sure anyone who might be unpleasantly surprised by what you find is in the loop. You can't prompt an LLM not to embarrass your boss, that's still up to you. If you are the boss, have at it. Not everyone is on the retro train. Steve Furber, co-creator of the BBC Micro and the Arm processor, has said he can't see the point of old computers because new ones are so much better. As he's spent his latter years designing million-core artificial brains, he has a point. But anyone who must keep alive that which should be dead – or put it out of everyone's misery – should watch what the older generation is doing with the corpses in its labs. It can be… electrifying. ®

AI is becoming both an attack tool and a high-value target, with attacks by AI-enabled adversaries rising 89 percent in 2025, according to CrowdStrike. The security firm's annual Threat Hunting Report details criminal gangs and nation states using AI throughout the attack chain. Attackers are also targeting organizations' AI infrastructure and poisoning popular software packages to compromise their users. "AI is both the weapon and the target," CrowdStrike counter adversary division senior VP Adam Meyers told reporters. "AI is a high-value attack surface, and it's being used by more and more threat actors." These attacks include LLMjacking, in which criminals steal corporate credentials to access frontier-model APIs, and cost harvesting – deliberately inflating a victim's AI usage to run up its bill. In one campaign, CrowdStrike documented a token thief sending about 200,000 API requests in just two minutes. The security vendor's threat hunting team now tracks AI agent-triggered leads at 2.5x the rate of human-triggered threats, and Meyers said this increased volume remains true across both government-backed goons and financially motivated criminals. CrowdStrike tracks more than 290 adversary groups, having added about ten this year. Of the 290, a North Korean crew it tracks as Famous Chollima – a sub-unit operating under the Lazarus Group umbrella and best known for its fake IT worker scams – "demonstrated the most advanced AI usage" over the second half of 2025 and first half of 2026, according to the report. This government-backed crew created "entire fake companies with AI-generated websites, GitHub accounts, and email infrastructure to support insider threat operations," the authors wrote. AI supply-chain compromise was the second most common MITRE ATLAS technique used by attackers to gain initial access, and Famous Chollima's campaign targeting AI-focused development environments "was one of the most sophisticated examples of this technique in practice," the report noted. This included a supply-chain attack in January and February targeting cryptocurrency and blockchain companies. In these attacks, the Norks published trojanized repositories, primarily hosted on GitHub, that contained legitimate-looking project files alongside hidden, malicious scripts. When developers opened these repos, malicious scripts automatically executed commands that gave Famous Chollima access to their environments. "AIs themselves are being targeted through that supply chain and through the CI/CD pipelines that they're dependent on," Meyers said. Threat hunters suspect another Lazarus Group offshoot, tracked as Stardust Chollima or Sapphire Sleet, was behind the March Axios supply chain attack. Last week, Amazon attributed four npm compromises over the past 18 months to the same North Korean crew. Meanwhile, a financially motivated crew tracked by CrowdStrike as Altered Spider and elsewhere as TeamPCP targeted developers' AI tools, compromising more than 300 software dependencies in one day. It harvested credentials and secrets before pivoting into cloud environments for theft and extortion. Altered Spider "hits the endpoint in seconds and within minutes, they're inside of the cloud," Meyers said. "It gives you a sense of how quickly they can move throughout that environment, and this is all tied again to software supply chains." Patching window slams shut CrowdStrike argues that AI is helping attackers exploit newly disclosed vulnerabilities at machine speed. From January to June, 88 percent of the exploitation observed by CrowdStrike using public proof-of-concept (PoC) code occurred within 48 hours of the code's release. The company said China-linked groups such as Vault Panda and Genesis Panda moved even faster, launching attacks within 24 hours of disclosure. "Vulnerabilities are weaponized through the use of AI," Meyers said. "This is creating a rich ecosystem of vulnerabilities for attackers to use against various systems, and what this really means is that the 30-day patch window, which frankly, was aspirational, is completely obsolete. We're down to 24-hour, 48-hour patch cycles, and organizations are really struggling under that." Meanwhile, as anyone who follows Microsoft's Patch Tuesday – or any other software vendors' vulnerability disclosures over the past few months – knows, AI is also really good at finding bugs in code. This means more CVEs and more patching for sysadmins racing to fix flaws before miscreants reverse-engineer the updates and develop exploits. "In 2025, there were something like 48,200 CVEs that were registered," Meyers said. "We're already, as of last week, at 43,000 for this year. We're not even into August yet, and we're already coming very close to the number from last year." June alone saw more than 7,600 software bugs reported and tracked through CVEs, he added. "The vulnerability ecosystem is going to be the big story for the next couple of months." ®

WHO, ME? Comedian W.C. Fields is credited with the aphorism "Never work with children or animals," which The Register mentions because half of it has some relevance to this week's edition of "Who, Me?" – the reader-contributed column chronicling the ups and downs of working in tech. This week, meet a reader we'll Regomize as "Stockley" who shared a horror story that took place in the 1990s. At the time, Stockley worked for a company that ran an IBM RS/6000, one of Big Blue's brood of Unix machines. "The business was run on BASIC, with code going back 25 years," Stockley wrote. When much of that software was written, memory was measured in kilobytes. "Programs had to be broken into chunks, which were called sequentially with the CHAIN command," he told Who, Me? The company's tech may have been a bit crusty, but it made up for that by being a lovely family-friendly workplace – so much so that it invited staff to bring their kids in for Halloween trick-or-treating. Stockley told us that one Halloween, the company's IT boss knew the children would disrupt work, so started the day's data processing early, setting the chained programs in motion. Then his wife and children arrived ahead of the Halloween festivities. The boss therefore parked his children in his office, in front of terminals connected to the RS/6000. The terminals were switched on and logged in as root. In Stockley's telling, the boss's nine-year-old daughter had recently seen a movie in which a young girl somehow acquired Unix skills and saved the day by typing "SHUTDOWN." And when the boss returned to his office, his daughter greeted him by saying: "Look, Daddy! SHUTDOWN worked just like it did in the movie!" Stockley recalls that the boss needed 36 hours to repair the damage his daughter had caused. Have kids messed with your tech? If so, don't mess about! Click here to send your story to Who, Me? so we can prove the point about working with kids on a future Monday. ®

Microsoft has added four new items to the to-do list it set itself to improve the quality of Windows 11. Redmond’s list landed in March 2026 in response to users’ anger at Windows 11 becoming increasingly flaky and needy. Pavan Davuluri, Microsoft’s executive VP for Windows and Devices, promised “Improved memory efficiency, lowering the baseline memory footprint for Windows, freeing up more capacity for the apps you run.” Last Friday, Davuluri reviewed Microsoft’s efforts in a new post that is unsurprisingly enthusiastic about Redmond’s repair job – even though much of the work is yet to appear in production versions of Windows. Among the work Davuluri lauded was removing some AI features, improving driver quality, and trying to make the Windows Insider program less convoluted. But he also acknowledges there’s more to be done and that memory efficiency is yet to manifest. “We’ve steadily been introducing several improvements to the memory efficiency of Windows, from taking advantage of a more efficient memory allocator to reduce overhead across apps and components, continued tuning of WinUI 3 so that apps built on it use less memory by design, as well as driving efficiencies across Chromium and Webview2 components when they appear in the operating system,” he wrote. Davuluri’s post also reveals that Microsoft has decided to work on four new areas it hopes will make Windows less craptastic. The third of those is “Memory optimization for 8GB and above,” which Davuluri describes as “Reducing Windows memory footprint to deliver a fast and responsive Windows experience across the PCs customers use every day.” Microsoft’s system requirements for Windows 11 state that the OS needs a PC equipped with at least 4GB of memory, or 16GB of pricey DDR5/LPDDR5 to qualify as a Copilot+ PC capable of handling AI workloads. The software giant seemingly assumes Windows users are all clamoring to run Copilot and other AI, so working to ensure that Windows 11 can run on 8GB of memory feels like a tacit admission that Copilot+ PCs aren’t exciting buyers, perhaps due to the current high price of memory. Or perhaps because people just don’t see much value in paying extra for an AI-capable PC. Another reason for the change may be the declining PC market, which analyst firm IDC recently forecast will experience a 11.3 percent shipment slump in 2026, and 20 percent drop in calendar Q4 alone. “The culprit is a persistent memory shortage with no meaningful relief expected before the end of 2027,” the firm stated. “The knock-on effects are significant: prices are rising and PC manufacturers are struggling to maintain full product portfolios.” Microsoft’s three other new Windows improvement priorities are: Delivering a faster and more efficient out-of-box experience. Providing a faster set up and making it easier for families to access useful parental controls. Making voice more natural and fluid to interact across the apps you use every day. Davuluri didn’t say when Microsoft will deliver these changes. But he did say the company has heard Windows users’ frustrations and will act to assuage them. “The signal from you is clear: keep going. We intend to,” he wrote. ®

This post contains major spoilers for House of the Dragon season three, episode seven. If you want to avoid them for now you can instead check out our coverage of episode six.

HBO ⚡ Quick Take

• House of the Dragon’s latest fake out was its most dramatic, but it was also its least surprising for a lot of obvious reasons.

House of the Dragon‘s third season has teased the death of Tyland Lannister, Aemond Targaryen, and Corlys Velaryon (twice). All of them turned up alive, but none with the panache of the show’s deus ex dragon-a return of Aegon’s Sunfyre. You know, the not-dead dragon who arrived exactly when his rider/the plot needed him.

HBO

House of the Dragon foreshadowed its Sunfyre fake out earlier in the season. Aegon visited his seemingly dead dragon at Rook’s Rest in episode four. That’s where Aemond bathed his brother in dragonflame, causing the King and his beautiful golden mount to crash into the woods below. Since then locals had started charging a small fee for people to look and touch the dragon, which was not breathing or generating heat/smoke, let alone fighting back. Aegon wasn’t convinced. “He’s alive. I know it. He’s alive. He must be,” he told an unconvinced Larys. No one seemingly had any reason to believe a desperate Aegon, but riders do have a special bond with their dragon, so Aegon couldn’t be dismissed entirely.

Everything we saw during that scene was a creation of the show. It was a big departure from the history told in Fire & Blood. Sunfyre never died at Rook’s Rest. He was greatly harmed and unable to fly in both versions, but despite his injuries, the Sunfyre of Fire & Blood was definitely not dead. Green soldiers brought it animals to eat after the dragon finished eating all the dead soldiers. It could also still breathe fire. Sunfyre used its tail and flame to protect itself from Rhaenyra supporters when they came to slay the stationary dragon. Those attempts ended when Sunfyre disappeared without a trace.

HBO

On House of the Dragon, Sunfyre left Rook’s Rest and showed up at the very last possible moment to save Aegon from the glorious death he wanted. It sure was luck for Aegon….

The best reading of the scene is that the spiritual connection between dragon and rider drew Sunfyre to Aegon’s side in the King’s moment of need. The worst is that this moment was the series pulling off another empty spectacle driven by plot convenience rather than logic. But no matter which side of those extremes you lean to, if you watched Game of Thrones, you knew Sunfyre was not actually dead.

HBO

Don’t worry if you can’t remember why, we won’t spoil anything by reminding you. All we will say is that on Game of Thrones Joffrey told a story about an infamous moment involving Sunfyre. House of the Dragon and HBO have maintained canon within its own televised version of Westeros. That would have shattered if Sunfyre had really died at Rook’s Rest. The prequel has not yet reached the event Joffrey described.

If you remember that tale, Sunfyre’s death wasn’t much of a fake out because it was obviously fake. But even if you forgot about it or never heard Joffrey talk about the Dance of Dragons, this dragon “resurrection” wasn’t exactly surprising. And at this point in season three, why would any return surprise us?

Mikey Walsh is a staff writer at Nerdist. He’s trying to trademark “Deus ex Dragon-a.” You can follow Mikey on Bluesky at @burgermike. And also anywhere someone is ranking the Targaryen kings.

The post HOUSE OF THE DRAGON’s Dramatic Fake Out Wasn’t Surprising At All appeared first on Nerdist.

This post contains major spoilers for House of the Dragon season three, episode seven. If you want to avoid them for now you can instead check out our coverage of episode six.

⚡ Quick Take

• House of the Dragon’s big dumb dragon fight in season 3’s seventh episode exemplified the show’s problem of preferring spectacle to logic.

HBO

I think House of the Dragon has abandoned its source material and ruined incredible moments by making them significantly worse. But, honesty, those changes are not my primary problem with the show. Some of my favorite adaptations are very different takes on stories I have loved my entire life. No, my main problem with House of the Dragon is that it’s a bad television show entirely on its own merits.

As someone who loves Westeros what’s especially frustrating is that it’s bad for all many of the same reasons Game of Thrones‘ last two seasons were total disasters. When characters aren’t running in place in plots that go nowhere, they behave like idiots in service to nonsense plots designed to generate grand spectacles and “shocking” moments devoid of logic and reason.

It’s all style, negative substance. And that was once again on full display in season three’s impossibly silly seventh episode. It was embarrassing installment whose failures are best exemplified by a mind-numbing, brain-breaking, totally absurd dragon fight.

HBO

It’s only fair I start this screed by giving the show credit for two things I think made sense from its otherwise impossibly idiotic dragon showdown. The first is that it makes complete sense Rhaenyra would finally get on her own dragon and fly off to deal with an issue. It involved her formerly disloyal husband possibly conspiring with her son’s killer. If anything was going to finally get Rhaenyra to act like a dragonrider, that was it. It also made sense the Queen left Baela and Moondancer behind to look after King’s Landing, especially when no one knows where Vhagar is.

So far, so good. And thus ends our use of the word “good.” Everything that happened after Rhaenyra and Addam (too easily) found Daemon and Rhaena was ridiculous. This sequence was the dragon equivalent of Ramsey Bolton’s hounds only emerging at the exact moment Sansa Stark was done taunting him. (Those starving dogs sure did have dramatic flair!) Only, somehow, House of the Dragon‘s scene was even dumber.

Sheepstealer is a wild, nearly uncontrollable dragon. It has only ever bonded with one living thing in the entire world. It had encountered Daemon and Caraxes already, so it’s not unreasonable it stayed calm when they met again. But you can’t defend that Sheepstealer didn’t even move, let alone protect Rhaena, when she ran off in fear. It didn’t even instinctively fly after her. Was Sheepstealer just confused at first? Maybe. And maaaaybe all that time spent alone with Rhaena in a cave, even without any actual training, made him slightly more domesticated. But we can’t give the dragon/the writers that benefit of the doubt because of what immediately happened. The ornery, antisocial, untrained dragon just sat there patiently even when Addam grabbed Rhaena against her will and carried her back.

HBO

I know it seems like I am complaining about the actions of a fictional creature, but what I am really complaining about is that this show is more concerned with “big moments” than it is internal logic. And I know because of everything that happened after Sheepstealer inexplicably sat idly by while the only living thing it cares about was taken prisoner.

Once the show remembered a bunch of angry dragons should be, you know, angry, the fire started. Syrax attacked Sheepstealer. Then Seasmoke jumped in. Those two experienced, fully-trained dragons should have easily killed the clearly overmatched Sheepstealer. They had him pinned down by the neck! But then Daemon ordered Addam to follow the Queen to safety while he and Caraxes finished the job. It should have been very easy for Daemon to do that. Caraxes has as much battle experience as any dragon on the planet and Sheepstealer was already half-dead already.

So of course Daemon easily got tossed off Caraxes back so the weakened Sheepstealer could get away.

HBO

Now, obviously Daemon got right back on his healthy dragon and give chase to the limping, greatly wounded Sheepstealer, right? He didn’t let the tiniest setback in history stop him from finishing the job? There is no reason for him to give up.

……….You fool. You complete dolt. Why would he do something so logical? Why would he do the most obvious thing in the world? How can you even suggest such internal logic when, by letting Sheepstealer escape, the dragon can now make a dramatic return creating yet another big moment!

Did you not see the totally-not-dead Sunfyre’s glorious deus ex machina-like return at the exact moment the plot needed him to show up? (In an entirely original scene the show created from whole cloth.) Wasn’t that super cool and not super dumb? Isn’t that what makes Westeros awesome? Not smart writing that combines thoughtfully developed characterization, smart plotting, rich ideas, and wildly entertaining action and humor, but spectacle! The kind you can easily create without thought or concern or care or logic or anything that remotely makes sense. That’s the kind of writing that made the end of Game of Thrones so beloved!

HBO

My apologies. I know, that’s a whole lot of snark. I had to get that out of my system. It was vital because when Sheepstealer makes his inevitable, cockamamie surprise return, I won’t have any of old frustration pent up. I’ll need room for the new frustration.

I promise I get why “big dragon fight” is a cool idea and on its own can be enjoyable even if its dumb. I’m not inherently against shows or movies doing things simply because they’re fun and entertaining even if they’re dumb. I love the Fast & Furious franchise. But that’s not what I want or expect from this franchise and I’m beyond tired that’s what it keeps giving me.

I don’t know how everyone who also loved Game of Thrones isn’t tired at this point. House of the Dragon—a story about a bunch of dragonriders in a civil war known as the Dance of the Dragons because it involved so many actual dragons fighting—did the seemingly impossible. It found a way to make a big dragon fight unwatchable and stupid. That sums up what this series has become since season one. Because the problem with this show is not that it’s a bad adaptation, it’s just bad.

Mikey Walsh is a staff writer at Nerdist. You can follow Mikey on Bluesky at @burgermike. And also anywhere someone is ranking the Targaryen kings.

The post HOUSE OF THE DRAGON’s Big Dumb Dragon Fight Highlights the Show’s Failure appeared first on Nerdist.

⚡ Quick Take

• In House of the Dragon season three’s seventh episode, Helaena has a series of sobering yet interesting dreams.

If I could help any character in House of the Dragon escape all of its madness to live a happy and healthy life, it would be Helaena Targaryen. She is the only daughter of Alicent, a woman who is willing to sacrifice her own kids and grandchild to save her own skin. She has two insane brothers, one of whom she was pushed to marry (yuck) and another one that she doesn’t even know. Helaena’s son was brutally murdered and now her rage and grief filled half-sister is holding her and her daughter hostage. Oh, and Helaena has no friends and she’s pregnant again. She’s a kind and pure soul who is caught in the middle of everyone’s mess, and I hate that for her. In fact, Helaena cannot even catch a break in her dreams. 

Theo Whiteman/HBO

In House of the Dragon season three’s seventh episode, Helaena has a series of really sad yet interesting dreams that speak to her state of mind. After Rhaenyra harasses her to no avail, Helaena dreams of giving birth to a baby boy, who is swiftly taken away from her before she can even truly see or hold him. Then, we see her in full battle wear as she mounts a dragon and burns an entire army of men. The next scene is her with her twin children (who are now older) living in peace and feeding chickens, a reality that she will never get to experience. The sky begins to darken and it starts to snow before a cold mist takes over the land. 

While there could be some deeper future meaning into these visions, they appear to be pretty straightforward. She hopes that the birth of her child won’t lead to his death and her becoming a brutal and cold person like other Targaryens. And, even though she dreams of a life in the light, she knows that winter is coming and this will not end well for anyone. It’s a sad life for someone to live, and her own mother is to blame. Just Helaena, I also question why she didn’t just let Helaena marry off into some wealthy family and get away from all of this.

RELATED ARTICLE

Every Major Character Death in HOUSE OF THE DRAGON Season 3

Sure, there was a chance that her husband could be a not-good man. In fact, most men in Westeros are not great husbands. But it could have been better than marrying her brother and becoming a pawn in this, ahem, game. Alicent even admitted that part of the reason was her own selfish desire to keep her daughter close. As a mom, it makes sense to want to protect your daughter but you know that your own son is a horrible man.

What will happen to Helaena in House of the Dragon? Probably nothing good, I’d guess. If this child is a boy, it is not beyond reason to think that, instead of becoming the person she did in her dream, that she would rather end her own life. We will have to see what else Helaena’s dreams reveal as this season starts to come to an end.

The post HOUSE OF THE DRAGON’s Helaena Has a Series of Sad Dreams appeared first on Nerdist.

ASIA IN BRIEF Meta last week briefly took down an unremarkable video posted by India’s prime minister Narendra Modi, earning itself days of criticism and regulatory trouble. The video featured Modi announcing a new task force that will reform the exams required to secure places in many Indian universities. Those exams have become a major factor in the sudden development of a major youth protest movement in India, after the cancellation of one test amid claims that questions had leaked. As part of his attempts to quell protests, Modi posted his first-ever selfie-style Instagram reel last week and followed it up with others –including the one Meta took down and replaced with a placeholder saying the vid was the subject of a legal complaint. The social networking giant said that was a mistake, apologized, and restored the video. That error gave India’s government the chance to beat up Meta in public. S. Krishnan, the secretary of India’s Ministry of Electronics and Information Technology (MeitY), said he asked Meta “to come in at the highest level and explain what is happening and why.” “They have agreed to come – well, they have to – and explain what the situation is,” Krishnan said. “We want to have both a policy level understanding and a technical understanding of the issues, and also adaptation to the kind of concerns India has.” Krishnan said Meta has written to India’s government to express regret over the incident and established new protocols governing moderation of accounts run by prominent people to avoid future messes of this sort. India has a complex relationship with Big Tech companies, sometimes celebrating their role in assisting the nation’s development, but often also lamenting their monopolistic tendencies and role in spreading content the Modi government would rather netizens don’t see. DeepSeek teases ‘peak/valley’ pricing for new models which hit beta last week Chinese AI upstart DeepSeek last week released a beta API for its next model, deepseek-v4-flash, and revealed a new pricing policy for its APIs. The company says the new policy will apply “soon” and will see the cost of its services double between 09:00 and 12:00, then again from 14:00 to 18:00 – in China’s single official timezone UTC+8. DeepSeek startled the AI market last year by releasing models that it claimed to have trained without vast fleets of Nvidia accelerators. Those claims didn’t stand up to close scrutiny, but the company is considered a significant challenger to western AI outfits such as OpenAI and Anthropic. deepseek-v4-flash is likely to reach general availability within weeks. DeepSeek already claims it outperforms its last model, V4-Pro. Australia proposes new tax on Big Tech – and a new out Australia’s government has tweaked its plans to tax Big Tech companies unless they pay local media for the right to share links to their work. The Land Down Under previously threatened tech companies with a 2.25 percent tax if they don’t fund local media. On Monday, that plan changed to a 2.5 percent levy on advertising revenue only and expanded incentives to fund small and regional publications. Meta has previously slammed the idea as “A discriminatory tax built on a false premise.” Also in Australia, regulators last week decided to sue Telegram after it failed to take down violent and terrorism-related content. Kioxia profits jump by billions, as you’d expect from a memory-maker Japanese memory-maker Kioxia last week announced [PDF] its quarterly revenue grew 451 percent year over year to reach $11.1 billion, plus profit of $5.3 billion – a massive turnaround from the $287,000 loss for the same quarter last year. According to Japanese outlet Nikkei, CFO Yoshihiko Kawamura said the company expects even better days ahead because he believes demand for NAND storage is “still in the early stages.” Japan’s next moonshot to ride local rocket Japanese space startup iSpace last week announced its next mission will ride a Japanese rocket to the moon. iSpace has launched two moon lander missions, but both failed. The company’s plan for its third mission involves new lander design called ULTRA that is capable of carrying payloads weighing several hundred kilograms to Luna. This time around, iSpace plans to use the H3 rocket built by Mitsubishi Heavy Industries, instead of the SpaceX Falcon 9 that carried its first two missions. ®

Survey Says is a weekly series rounding up the most important polling trends or data points you need to know about, plus a vibe check on a trend that’s driving politics or culture. In this year’s midterm elections, the hardest Senate seat for Democrats to defend probably isn’t what they expected a few months ago. Polls show Georgia Sen. Jon Ossoff handily leading Republican challenger…

Source

A cartoon by Mike Luckovich. Related | Who will take the fall for Trump’s Iran war fiasco?

Source

ph @casanova_coffee_art
Gelateria Quadrifoglio Noale

Under an intensifying trade war, parts of northern Michigan are feeling the financial pain. But with a president intent on conflict and a congressman silent on the issue, these Michiganders see no sign of relief. By Anna Clark, photography by Emily Iris Elconin, for ProPublica On the northeast edge of Michigan’s Upper Peninsula, nearly 1,100 people gathered in late June for the…

Source

The global memory chip shortage appears to be affecting the availability of Apple’s most popular Mac.

By Tony Leys for KFF One of America’s most competitive gubernatorial races could settle a heated argument over whether private insurance companies should run Medicaid. The race is in Iowa, whose Medicaid program has been plagued with controversy since 2016, when the state hired national insurance companies to manage billions of dollars’ worth of benefits. That shift was made by…

Source

On the latest episode of Equity, we discuss why Sam Altman has calling on the industry to "pace the rate of AI development."

Congressional Cowards is a weekly series highlighting the worst Donald Trump defenders on Capitol Hill, who refuse to criticize him—no matter how disgraceful or lawless his actions. Senate Republicans on Wednesday dragged Anthony Fauci in front of the Senate Homeland Security Committee and berated him for hours over his handling of the COVID-19 pandemic, accusing him of trying to…

Source