In this episode, Ray Cochrane leads with Nvidia’s board investigating reports that Chinese intelligence targeted CEO Jensen Huang. The same day, Huang told employees he had been unfaithful to his wife. Additionally, he covers the Justice Department’s Super Intelligence Task Force and the fake job interview behind Binance’s $6.3 billion crypto Heist. He also gets to a Sega hack that darkened arcades across Japan and the fall of NASA’s Swift telescope.

– Want to start a podcast? It’s easy to get started! Sign up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Ray Cochrane opens with a quiet personal update. He had a relaxing weekend, spending Friday and Saturday night with Grandma and Delaney. They went out to dinner together and just hung out.

In his free time, Cochrane is building a new podcasting tool in Rust. So far, the recording API works, along with some diagnostics and debugging tools. After so much time in PHP, he calls it a breath of fresh air.

On the audio side, he noticed the mic was still picking up some feedback last episode. He hasn’t rewired the studio yet, and he isn’t sure which device causes the static. Meanwhile, he is adjusting settings and slowly working through the knobs while keeping the voice quality up.

Nvidia’s Board Investigates Its CEO

The lead story comes from CBS News. On November 8, it reported that Nvidia’s board opened an investigation involving its own CEO. Nvidia designs the chips that much of the AI industry runs on. The board says it is examining news reports that co-founder and CEO Jensen Huang “has been the target of a foreign intelligence effort.”

The same day, Huang sent employees a letter, which Nvidia posted on its news site. In it, he admitted he was unfaithful to his wife and apologized. The letter names no one. Its only nod to the spy reports points to the board’s independent investigation, which Huang says he supports.

Independent directors, board members who aren’t company executives, will lead that investigation with outside counsel. So nobody who reports to Huang is running it, which Cochrane counts as good news. The board adds that it has “no reason to believe that either national security or NVIDIA secrets were compromised.”

Huang stays on as CEO. From what Cochrane has heard, most people like working under him, so he finds the news tragic.

How the Story Broke

The story broke on October 29 in The New York Times. The paper reported that a businesswoman named Alice Huang had grown close to Jensen Huang. According to the Times, US intelligence officials had concluded she was passing information to Beijing. Despite the shared last name, the two aren’t related.

The Times described her as an executive at Megaspeed, a Singapore company that buys Nvidia chips. It also said President Trump was briefed about a year before the story ran and “chose not to do anything with the information.” Later, TechCrunch, citing the Times, reported on a US government memo from earlier in 2026. According to that memo, China’s Ministry of State Security, its main spy agency, recruited her to watch Jensen Huang.

On November 6, the Financial Times reported that the same memo describes the relationship as romantic. Then the San Francisco Standard, also citing the Times, reported that Huang named Alice Huang to the board as the other person in the relationship. To Cochrane, Huang seems to be trying to be as transparent as possible.

However, no charges have been reported against Alice Huang, and she hasn’t commented. The spy claim rests on unnamed people the Times says were briefed on the surveillance. Meanwhile, China’s foreign ministry called the reports “baseless and purely slanderous smears against China,” which Cochrane calls the response you’d expect.

Nvidia, for its part, told ABC News that Huang’s interactions with her were “not unusual or unexpected.” In October, Huang also told Bloomberg News his contact with her was “related to Megaspeed, and it was all completely normal and not unusual.”

Megaspeed and the Chip Diversion Worry

Megaspeed runs data centers full of Nvidia hardware, and customers pay to use that computing for AI work. According to the Times, its Nvidia chip purchases topped $4.8 billion over about two years. CBS News counts it among Nvidia’s largest customers in Southeast Asia.

US officials have looked into Megaspeed since 2024. It started when a Commerce Department official noticed Nvidia equipment at a Malaysian data center that had never been unpacked. Then, in October 2025, Senator Tom Cotton of Arkansas called for a Commerce Department investigation of the company.

The suspicion is chip diversion. Export controls are the US rules on which technology can leave the country and who may buy it. Under them, Nvidia’s most powerful chips can’t legally be sold into China. In a diversion, chips go to an approved buyer in another country, then move on to a barred one.

Cochrane can see why China might try it, since it wants the best chips available. In an unrelated case, federal prosecutors in Los Angeles charged a California man on October 1 with smuggling more than $300 million in export-controlled servers to China. He hasn’t been convicted.

Chips can also serve Chinese customers without ever entering China. As Forbes describes Cotton’s concern, Nvidia chips sat in Malaysia and Indonesia while Chinese customers used them from a distance. Investigators also suspect Megaspeed was a route for some chips into China itself.

Megaspeed hasn’t been charged with anything, and it says it hired outside counsel for an internal investigation. Nvidia told CNBC that “the diversion allegations against this regional cloud were thoroughly investigated and debunked last year.” Still, the Times reports the US investigation moved to the Justice Department in 2026.

The H200 and China

In January, the US began allowing one older high-end Nvidia chip, the H200, to be sold to approved Chinese buyers. CNBC reported the US government takes 25 percent of those sales, while newer chips stay barred. Even so, Chinese regulators have kept the H200 out. An Nvidia spokesperson told CNBC that China’s government “blocked U.S. AI chips, protecting its own industry.”

Cochrane finds that an interesting approach. Still, he suspects China would study the chips it already has to compare techniques.

Huang has argued for selling more to China, and CNBC reported he is close to Trump. According to the Times, he met Trump in April 2025 to argue against the ban. Officials had two worries, the Times reported: what could leak out of Nvidia, and how much access Huang had to the president.

When US officials believe a foreign spy service is targeting an American, they can warn that person in a private briefing. Yet the Times reported in October that Huang never got that warning, which Cochrane finds interesting.

Congress Wants Answers

On November 4, three Republican House committee chairmen asked Huang to a closed-door briefing on November 18. Their committees cover foreign affairs, intelligence, and China. The briefing is a private, voluntary meeting with lawmakers. Over the weekend, Democratic Senator Chris Coons of Delaware said, “Congress will investigate and get answers.”

The briefing lands one day after Nvidia reports quarterly earnings on November 17. As of November 8, the company hadn’t said whether Huang would attend. Additionally, the board hasn’t named its committee members, its outside counsel, or a timeline.

Cochrane calls it a big story out of Nvidia, and a sad one. People get caught doing the stupidest things, he says, and this is one of them.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

The Justice Department’s Super Intelligence Task Force

The headlines open with a story Cochrane found a little wild. On November 5, the Justice Department announced its new Super Intelligence Task Force. The name traces to a September 29 executive order. It tells federal agencies to write “super intelligence” where they used to write “artificial intelligence,” as far as the law allows.

So AI becomes SI, though Cochrane jokes that many might agree it isn’t very super. The order defines SI as the same technology federal law already calls AI. Agencies don’t have to change earlier regulations or contracts. Meanwhile, the president’s science advisor owes him draft language for a new legal definition by November 28, and only Congress can make it law.

That same September day, Trump and six tech leaders signed a voluntary pledge called the White House Accord on Super Intelligence. According to CNBC, one of them was Jensen Huang, from the lead story.

Attorney General Todd Blanche created the task force in an October 13 memo. Its targets include fraud by people and companies, international crime rings, and foreign governments, along with the groups they back. It coordinates cases from existing offices. It also takes over a unit from January that challenged state AI laws in court, The Next Web reported.

There’s no word yet on staffing or cost. Tech Times also noted that prosecutors still need laws that existed before September 29. Meanwhile, PolitiFact, writing on Poynter, checked what the word means inside the AI industry. It wrote, “Superintelligence is AI that exceeds human cognition in nearly all fields, something the industry generally agrees no system has achieved.”

Senator Mark Warner of Virginia, vice chairman of the Senate Intelligence Committee, said: “Changing what we call artificial intelligence does nothing to address the very real risks that come with increasingly powerful AI systems.” Cochrane calls it a loaded change that may not even go through, and he plans to revisit it if it becomes official. Maybe, he muses, they know something the public doesn’t.

Google’s $17 Billion Wyoming Data Center

Next comes a callback to episode 1883, whose lead story covered voters turning down data centers. On November 5, Google said it will spend $17 billion in Wyoming over the long term, starting with its first data center in the state. Yet WyoFile pointed to a University of Wyoming poll from late October. In it, just 30 percent of residents said more large AI data centers should be allowed.

County filings, reported by Cowboy State Daily in July, put the campus in Laramie County, about 8 miles south of Cheyenne. Those filings named Google as the owner of a project the county approved in January under another developer. They show four data centers, with a target completion date of 2031.

Google says its first power source will be 702 megawatts of Bloom Energy fuel cells running on natural gas. Fuel cells make electricity through a chemical reaction instead of burning fuel, but they still give off carbon dioxide. Google says most of it will be stored deep underground nearby. Gas generators from Tallgrass and the utility Black Hills Energy will supply more power, and Google’s post doesn’t mention wind or solar.

On costs, Google wrote, “We will pay for 100% of the power we use, and when we build data centers, we’ll cover the infrastructure needs directly driven by our growth.” Cochrane thinks that’s the only way communities will start allowing data centers. He points back to last episode’s county, where data centers covered a big share of the budget on a small share of the land. Covering all costs makes the deal enticing, which he finds mildly concerning.

Cheyenne Mayor Patrick Collins told WyoFile, “This is not speculative. This is happening.” City Councilman Larry Wolfe told the same outlet, “You know, we’re not easily dazzled anymore.” WyoFile also reported that Wyoming Secretary of State Chuck Gray opposes more AI data centers there. As Cochrane notes, some people still object despite the huge sum on the table.

A Fake Job Interview Behind Binance’s $6.3 Billion Heist

BleepingComputer reported on November 3 that a phony job interview opened the door to the biggest crypto theft yet. On October 29, attackers moved about $6.3 billion out of the exchange Binance in under four hours. That’s more than four times the old record, Bybit’s $1.5 billion loss in February 2025. The FBI pins both on North Korea, through a hacking team it calls TraderTraitor.

Binance’s own report traces the attack to June, when a wallet engineer downloaded a tampered installer for an AI coding tool. Its security staff caught it and swapped out his laptop. Then, in September, a fake project director contacted a second engineer through LinkedIn, according to Fortune. Near the end of his second interview, he was told to download a coding test onto his work laptop.

Hidden code ran while he set it up. It was a zero-day malware no one had seen before, so security software had nothing to match it against. On October 20, the attackers used one of Binance’s internal AI tools to take over the June engineer’s computer and a third engineer’s. The June engineer’s machine held a digital pass for launching software on company servers.

On October 29, they used it to reach the system that approves moving coins out of Binance’s hot wallets, the ones kept online so withdrawals go out fast. Binance says about $2.85 billion has been frozen or intercepted, and that it, not its customers, will cover the loss. Its report says, “This should not have happened.” North Korea denies any role, and Cochrane isn’t buying it.

A Sega Hack Darkens Arcades Across Japan

Tom’s Hardware reported on November 8 that arcade games across Japan went dark for about a week after a break-in at Sega. The network belongs to Sega Fave, a Sega group company, which says an outsider got into its servers. On November 1, it shut down ALL.Net, the network linking arcade machines to its servers. The tap-card system players use to load saved progress went down with it.

Bandai Namco runs eight arcade games on ALL.Net, including Taiko no Tatsujin and Mario Kart Arcade GP DX, so those went dark too. When ITmedia News visited a Tokyo-area arcade on November 5, the network games were dark while the crane games kept running. Bandai Namco restarted its games for testing on the morning of November 8, Japan time.

However, Sega’s own games, including Initial D and maimai, were still down, with Sega’s next update due November 10. Payment details, including credit cards, sit on separate systems that weren’t touched. So far, Sega Fave hasn’t found customer or partner data leaving its servers, and it hasn’t said who did it or how. Cochrane hopes for a follow-up on the culprit and the motive.

Seafloor Mud Reveals Hidden Quakes Off Japan

Phys.org carried a November 8 report from a University of Tokyo team that includes Professor Yusuke Yokoyama. They went back to a 16-inch tube of seafloor mud pulled up in 2012 off Japan’s Kii Peninsula. Its deepest layer dates to around 1440. Yet the mud looked the same all the way down, whether checked by eye or with X-ray and CT scans.

When an underwater slope collapses, it scoops up seafloor mud and drops it all at once, leaving what researchers call an event layer. Yokoyama explained, “Because the redeposited sediments are the same as the materials below, the chemical and structural features were indistinguishable, making the event layers invisible.”

So the team measured carbon-14 at spacings under half an inch. Since undisturbed mud gets older with depth, older carbon sitting on younger mud means something stirred it up. Tiny shelled creatures in the same mud gave them calendar years.

The Nankai Trough is where one of the plates that make up Earth’s surface dives under southwest Japan. Yokoyama said, “We found several new geological layers that were not visible before, derived from submarine landslides possibly triggered by earthquakes in the Nankai Trough.” Floods and storms are other possible causes. He added that the team could “correlate some of these with written historical records of submarine earthquakes.”

Japan’s government puts the odds of a magnitude 8 or 9 quake there within 30 years at 60 to 90 percent or more by one method. By another, it’s 20 to 50 percent. Cochrane can never make sense of that range. Realistically, he says, it means a 20 to 90 percent chance, which hardly works as a metric.

AI Tells Individual Birds Apart by Song

Science reported on November 6 that an AI system can now pick out individual ovenbirds from recordings of their songs. A team led by Sam Lapp of the University of Pittsburgh published the work in Nature Methods on November 5. On birds it had never trained on, the system named the right bird 96 percent of the time.

Right now, tracking which birds survive each year means leg bands. Every bird has to be caught first, which stresses it and takes time. Bands also get lost before anyone checks the next year. The test set came from Alberta, Canada, where microphones had already tied each song to a known bird.

Male ovenbirds repeat one song, and no two males sing it exactly alike. The AI learned those differences and grouped the songs so each group could be one bird. After that check, the team followed 405 males across 126 sites over four years. Biologist Scott Sillett of the Smithsonian Migratory Bird Center, who wasn’t part of the study, told Science, “This is almost a holy grail.”

Still, the method only counts males, since they’re the ones singing. Lapp says the hardest part for other species is ground truth, meaning recordings where someone already knows which bird sang each song. Handling birds also gives researchers health information a recording can’t. As Lapp told Science, “The human still needs to be a part of the process.”

Cochrane takes that to mean today’s researchers won’t lose their jobs, and he calls the work awesome.

NASA’s Swift Telescope Falls to Earth

Next comes a sad end to a story the show followed in episodes 1869 and 1871, as NASA attempted to rescue its Swift space telescope. CBS News reported that Swift fell back into the atmosphere at 7:39 a.m. Eastern on November 4, over North Sumatra, Indonesia. It had been in orbit for nearly 22 years.

Swift launched in 2004 on a mission meant to last two years. It hunted gamma-ray bursts, brief flashes of the most energetic light, which can come from collapsing giant stars and colliding neutron stars. Other observatories counted on its alerts, since it could swing its telescopes onto a burst quickly. By NASA’s count, it observed more than 2,000 bursts and 1,400 exploding stars, which Cochrane calls a great investment.

Years of heavy solar activity heated and puffed up Earth’s upper atmosphere, and the extra drag pulled Swift lower. So NASA paid Katalyst Space $30 million to send a robotic spacecraft to grab it and boost it higher. On August 19, they called off the capture because they couldn’t reliably control the rescue craft. It never came closer than about 7 miles.

In August, NASA Administrator Jared Isaacman said, “This is not the outcome we were working toward, but it does not change why this mission was worth attempting.” NASA put the chance of falling pieces hurting anyone at about 1 in 2,100. Most of Swift burned up, and space writer Leonard David wrote that it wasn’t yet known whether any pieces hit land or water. Cochrane hopes nobody was hurt.

A Gas Over a Martian Volcano Hints at Magma

Smithsonian Magazine reported on November 6 about a gas over a Martian volcano that may come from underground magma. The gas is hydrogen chloride, which volcanoes on Earth can give off as rising magma loses pressure. An international team published the finding in Science Advances on November 4. It used readings from the European Space Agency’s ExoMars Trace Gas Orbiter.

Hydrogen chloride normally shows up all over Mars in its warm season. However, these readings came in the cold northern summer, in short spikes, and only over Alba Mons. That volcano covers about as much ground as the lower 48 states. The orbiter’s camera also spotted fresh-looking cracks on its northern side, each about 330 feet long.

The team thinks magma deep down gives off gas that seeps up through cracks like these. Ice seals the cracks in winter, and summer lets the gas out. While the paper calls ongoing magma activity the best explanation, co-author Kevin Olsen of Oxford told National Geographic that “the nature of the source is unknown.”

Hydrothermal systems are networks of hot water moving through underground rock. The study’s lead author told Reuters, “Such active hydrothermal systems may provide the energy sources and favorable conditions needed to support microbial life.” National Geographic reported that nobody expects an eruption soon. For Cochrane, active magma could open up many new hypotheses, and possible microbial life on Mars would be a huge research topic.

Cochrane wraps up with housekeeping. Listeners can become a GNC Insider at geeknewscentral.com/insider, email [email protected] to reach Ray or Chris, or leave a comment on the episode. He also points to podcastapps.com for a modern podcast app, and he uses Pocket Casts himself. Finally, he thanks GoDaddy for more than twenty years of keeping the show on the air.

He also admits the newsletter at geeknewscentral.com hasn’t been going out. He wasn’t sure how his dad handled it, but now that the show runs more consistently, he plans to restart it. Rather than a wall of article text, he’s thinking of showing links and a short personal update. Listeners with a preferred format are welcome to send it in.

The post Chips and Secrets #1884 appeared first on Geek News Central.

In this episode, Ray Cochrane leads with the server farm revolt. Voters rejected data centers in 40 of 41 local ballot measures on November 3, and two days later the Nasdaq had its worst day since April 2025. Additionally, he covers Anthropic’s IPO filing, a made-to-order melanoma vaccine, the finest Mars dust yet, and how voting is like a Klein bottle.

– Want to start a podcast? It’s easy to get started! Sign up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Ray Cochrane opens with Halloween. On Saturday, he went out in costume with Delaney, Zach, and Monique. Cochrane and Delaney went as a matching pair of Blue Man Group members, with blue face paint and drumsticks. The photos came out pretty hilarious, and he is still debating whether to share them.

He also owns up to the audio on the last episode, which he says was shot. That day, Cochrane spent five or six hours debugging the studio’s audio chain, and he hopes it is finally stable. One big problem was that his headphones didn’t match the exported file. The export came out far quieter than what he heard while recording.

Next, he plans to rewire the whole studio, hopefully this weekend, with masking tape labels on everything. Otherwise, life has settled into a slower pace, and he is loving it so far. Dinner that night was grilled venison, which he calls very tender.

All the audio work has also pulled him into how audio works under the hood. Lately, he has been playing with Rust and audio software while working through Rust in Action. Compared to other Rust books, he likes how well it leads readers along. So he recommends it to anyone curious about Rust.

Voters Turn on Data Centers

The lead story comes from CNBC, where MacKenzie Sigalos reported on November 4. Ballotpedia counted 41 local ballot measures about data centers in 10 states on November 3. In official results, data centers lost 40 of them. All 27 measures that set new limits passed, while voters turned down all 13 that would have approved projects.

Only Harwood, North Dakota went the other way. There, a proposal requiring expert review and a public hearing before big industrial projects failed 325 to 105. Ohio held 21 of the measures, and data centers lost every one. In Piqua, a ban went into the city charter with about 67 percent of the vote.

Piqua voters also recalled three officials who had backed the project. Voter Marty Pollock told Dayton’s WYSO the deal felt like it was done sneakily, under the radar. Cochrane said the voters did well to do their research, especially when leaders seem to work behind their backs with money on the table.

Data Centers in the Big Races

Data centers also came up in some big races. According to AP results, Democrat Gina Hinojosa beat Texas Governor Greg Abbott by about seven points. She had called for freezing new data centers until lawmakers write statewide rules. Meanwhile, in Ohio, Sherrod Brown unseated Senator Jon Husted after calling him “the face of data centers.”

Cochrane is careful not to credit those wins to data centers alone. Still, he sees the same sentiment against AI data centers nationwide.

An October poll by ABC News, The Washington Post and Ipsos found 72 percent of Americans don’t want a new one built in their own community. However, The Washington Post reported the issue still came in last of the five that NBC’s exit poll asked about. Cochrane makes a point of not conflating the two.

Power Bills and the PJM Auction

So why would a town vote against a building full of computers? Electricity comes first. Lawrence Berkeley National Laboratory found US data centers used about 4.4 percent of the country’s power in 2023. By 2028, it projects 6.7 to 12 percent.

Much of the eastern US runs on a grid managed by an operator called PJM. Each year, PJM holds an auction that pays power plants to stay available, and that cost lands on household bills. By IEEFA’s count, the price jumped from about $29 to about $270 in one auction, then to about $329. Cochrane calls it a crazy jump, roughly 11 times the old price.

PJM’s independent market monitor blamed data centers for 63 percent of that first jump. At a town hall in Prince William County, Virginia, Connor Pike said his family’s bill had climbed to $180 a month. Meanwhile, Ohio regulators now require big new data centers to cover at least 85 percent of the cost of the power they reserve.

Water and Tax Breaks

Water is the second fight. Many data centers cool their servers by evaporating water, and a large one can go through about 5 million gallons a day. During a drought in Fayette County, Georgia, a data center campus took more than 29 million gallons through meters the county wasn’t billing. The company, QTS, paid a back bill and blamed a county meter error.

Tax breaks are the third complaint, and Good Jobs First counts more than 30 states offering them. Moreover, The Register reported Virginia gave up about $1.6 billion in sales tax in 2025 alone. Yet a typical data center there runs on about 50 full-time workers once built, after as many as 1,500 during construction. Cochrane admits he had never looked at it from the tax side before.

Supporters point to Loudoun County, Virginia. Data centers sit on about 4 percent of its commercial land, but ITIF says they cover about 38 percent of the county’s main budget. Even so, Cochrane asks whether that money is worth the strain on land, water, and power prices, which he sees as the wider effect.

Wall Street’s Worst Day in Months

On November 4, the day after the election, the Dow jumped about 1,570 points. According to Yahoo Finance, no day after a midterm had been that good since 1982. Then, on November 5, it all came undone. The Nasdaq closed down 5.5 percent, its worst day since April 2025, and NBC estimated $2.8 trillion in S&P 500 value vanished.

Chip stocks led the drop. The Philadelphia semiconductor index fell more than 12 percent, its worst day since March 2020. Nvidia fell about 9 percent, while AMD and Micron each lost about 17 percent. Neoclouds, the younger companies that rent out computing power, fell hard too, with IREN down about 17 percent.

Cochrane doesn’t expect the big AI companies to slow down. However, he thinks smaller ones will get pruned as investment tightens, and that might not be a bad thing. He also agrees with Ed Yardeni that the AI bubble hasn’t burst yet.

AMD had just reported record data center revenue of $8.5 billion. Still, Cochrane finds it concerning that the company leans on that business to compete with Nvidia.

More Votes on the Calendar

More votes are already scheduled. On December 1, Clarksville, Arkansas, decides whether to keep city deals for up to $55 billion in data center financing. Then, on December 15, Mandan, North Dakota, votes on banning data centers that draw 25 megawatts or more. Cochrane notes 25 megawatts may be a fairly low bar, and he wants to look into it further.

Meanwhile, the energy data firm Cleanview looked at states whose governors have paused new data centers or promised to. By its count, those states hold over 40 percent of the data center capacity planned in the US.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

Anthropic Files to Go Public on the Nasdaq

The headlines start with Anthropic, the company behind the Claude chatbot. On October 26, it said it had publicly filed an S-1 with the SEC, the paperwork that comes before an IPO. Its shares would trade on the Nasdaq as ANT. CNBC reported revenue of about $16.3 billion for the first half of 2026, up from about $1.2 billion a year earlier.

An updated filing added about $17 billion for the third quarter alone. It also lists about $895 billion in spending commitments, nearly all of it for computing. That’s roughly seven times the cash and short-term investments Anthropic holds. Additionally, the risk section warns the company may never stay profitable, and that its fight with the Pentagon and the Trump administration could cost it revenue.

Cochrane thinks Anthropic made the right choice standing by its ethics in that fight, and he calls the backlash insane. Investors are split, and the Kalshi odds of a listing by November 30 fell from 84 to 60 percent in a single day. His advice to anyone thinking of buying in is simple: do your research.

Arm Explains What AI Agents Need From Chips

Arm laid out what AI agents need from computer chips in an explainer on its newsroom. Drawing on AI director Laurence Moroney, it breaks an agent’s job into four steps: figure out the goal, make a plan, act, then review. In Arm’s view, the chips, memory, software, and outside tools must all stay in step.

From an outside view, Gartner forecast in June 2025 that companies will drop more than 40 percent of their agentic AI projects before 2027. Cochrane can see why. As people hand more and more work streams to AI tools, he says, they lose track of what that work is doing. So he expects companies to refocus on their core products.

Arm also reported record quarterly revenue of $1.44 billion on November 4. However, Invesco reported its royalties rose only 8 percent, short of expectations, and the stock fell about 13 percent the next day. To Cochrane, that helps explain the drop, since Arm counts on royalties from AI data centers.

Amazon Builds One Front Door for Cloud Apps

Amazon Web Services announced ECS Gateway on October 20. ECS is Amazon’s service for running containers, apps packed with everything they need. Until now, teams built each app’s entrance by hand, with a load balancer, an HTTPS certificate, and firewall rules. With Gateway, an infrastructure team sets the rules once, and up to 25 apps can share one load balancer.

The Gateway itself is free, though customers still pay for load balancers and traffic. Cochrane sums it up as a reusable config, and a pretty cool one for anyone on AWS.

Meta Asks Small Business Owners About AI

Meta published a survey it paid for, run online by YouGov. Almost 4,000 small business owners answered across nine places in Asia and the Pacific. Of them, 84 percent said they use AI, and 76 percent said it made them more productive. However, those are answers, not measurements.

Meta’s own fine print says the results aren’t nationally representative, and PPC Land noted that Meta published no margin of error. By contrast, Singapore’s government found only about 23 percent of its small and midsize businesses had adopted AI in 2025.

Cochrane expects Meta to talk up its own survey. Still, he wishes it would publish place-by-place results with margins of error, which would help more than a list of AI’s upsides. He also believes Google offers a free AI course for anyone who wants training.

Google’s Classroom Game Has Kids Map a Brain

Google’s Polar Peaks is the third free game in its AI Quests series for middle schoolers, made with the Stanford Accelerator for Learning. Players imagine a character’s brain and trace its nerve cells to learn why it is losing memories. Along the way, they clean up messy data, train a model, and test it. Since there’s no chatbot or generative AI built in, Cochrane calls it pretty cool.

The game is modeled on Google’s work mapping connectomes, the full wiring diagrams of a nervous system. The biggest complete one maps a male fruit fly, with more than 166,000 neurons. A human brain, by comparison, has about 86 billion.

Meanwhile, Senators Richard Blumenthal and Josh Hawley sent Google’s CEO a letter about Gemini, which comes switched on by default on school devices. Cochrane says schools ought to turn that off, and he would be upset too.

Why Your Brain Is Wired Crosswise

Phys.org ran an essay by two University of Minnesota neuroscientists, Martin Wessendorf and Daniel Wang. It covers their new paper in Frontiers in Neuroanatomy. Much of the body’s wiring crosses over, which is why a stroke on one side shows up on the other. Smell and taste, however, don’t cross.

In the authors’ view, the senses that build a map of the space around you are the ones that cross. They trace that pattern back about 500 million years to the earliest animals with backbones. It’s a hypothesis built from published anatomy, and they suggest testing it with scans of ancient fish fossils. The paper also opens with a line from pitcher Bill Lee, which Cochrane calls a fire quote.

A Made-to-Order Melanoma Vaccine Passes Its Big Test

Science magazine’s Jocelyn Kaiser reported on a Merck and Moderna trial presented in Madrid on October 24. Each vaccine starts from the patient’s own tumor. Software picks up to 34 targets, which get written into one strand of mRNA. In the phase 3 trial of 1,137 adults, everyone got the drug Keytruda, and two-thirds also got the vaccine.

After about two years, the risk of the cancer returning or death was 42 percent lower with the vaccine. Cochrane clarifies what that means. About 82 percent of the vaccinated group were cancer-free at two years, compared with 70 percent without it. The vaccine isn’t approved anywhere yet, and the full paper is due in the New England Journal of Medicine on November 11.

Perseverance Finds the Finest Mars Dust Yet

Since landing in 2021, NASA’s Perseverance rover has carried swatches of spacesuit material on a calibration plate. Over 1,645 Martian days, it checked the dust on them with PIXL, its X-ray instrument. Teflon is made of carbon and fluorine, which PIXL can’t detect. So any signal from the Teflon swatch had to be dust, which Cochrane calls a clever approach.

Scott VanBommel of Washington University in St. Louis led the study. It found ultrafine grains about 100 nanometers, the smallest dust measured on Mars.

VanBommel says the smallest grains pose the greatest challenge for life support systems. Still, he told Reuters that doesn’t make Mars too dangerous for humans. Cochrane finds it terrifying to imagine breathing in dust that fine.

Europe’s CRISTAL Ice Satellite Opens Its Wings

The European Space Agency reported that CRISTAL, the Copernicus Polar Ice and Snow Topography Altimeter, has passed a round of factory tests. Airbus finished building it in Friedrichshafen, Germany, in September, then unfolded its solar panels for the first time. Next, it will spend most of next year in environmental tests at ESA’s site in the Netherlands. Launch is planned for late 2027 or early 2028.

Its key instrument is a radar altimeter, which times radar echoes to measure the height of whatever lies below. Snow sitting on sea ice muddies that reading. So CRISTAL uses two radar frequencies that react differently to snow, which lets it measure snow depth too.

How Voting Is Like a Klein Bottle

With the midterms in mind, Quanta Magazine’s Alex Stone wrote about Condorcet’s paradox from 1785. When three voters rank three candidates in a loop, each candidate beats another two to one. As a result, nobody wins every matchup. Later, in 1951, Kenneth Arrow proved no ranked voting system can follow a short list of fairness rules every time.

A new paper led by Ori Livson of the University of Sydney uses topology to glue the possible rankings into a Klein bottle. That’s a closed surface with only one side, and in this model the paradox shows up as that one-sidedness. It’s a preprint, and it claims no new proof or better voting system. Cochrane points listeners to the article for the visuals, since he keeps images out of the deck for copyright reasons.

Fire Guts the Eagle’s Nest

Smithsonian Magazine’s Soumya Sagar reported that fire gutted the Eagle’s Nest, the Kehlsteinhaus, high in the Bavarian Alps. The fire started on the night of October 23. Nobody was inside, and police say the building is largely destroyed. Hitler’s aide Martin Bormann had it built in 1937 and 1938 with Nazi Party money.

Cochrane keeps this one short. He finds it more shocking that it wasn’t taken down long ago. A Nazi home burned down, he says, and we should be happy about it.

Cochrane wraps up with housekeeping. Listeners can become a GNC Insider at geeknewscentral.com/insider or email [email protected]. He also points to the newsletter at geeknewscentral.com and modern podcast apps at podcastapps.com. Finally, he thanks GoDaddy for more than twenty years of keeping the show on the air.

Before signing off, he apologizes for the slide mix-ups during the show and promises a better flow. Hopefully, viewers will see an updated studio next time. He doesn’t plan to change much, since there’s a lot of history behind him, but the wiring must change.

The post The Server Farm Revolt #1883 appeared first on Geek News Central.

In this episode, Chris Cochrane tells the story of his dog Milo, who was paralyzed by intervertebral disc disease. He walks through the emergency surgery, the cost of care, and the long road of recovery at home. Chris also shares the GoFundMe he started to help with Milo’s medical bills.

– Want to start a podcast? Its easy to get started! Sign-up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Chris Cochrane opens with a personal reflection on a hard year. His father, Todd Cochrane, founded this show and passed away suddenly a little over a year ago. Chris and Ray took over the show after that. Chris says this past year has been one of the most difficult of his life.

This episode departs from the usual tech news. It centers on Milo, Chris’s dog, who became paralyzed a little over five weeks before recording. Chris walks through the diagnosis, the surgery, and daily life since then.

Meet Milo

Milo is Chris’s five-year-old dog and, in his words, his best friend. Chris got him around age 18, during the COVID-19 pandemic, when he needed a companion. Milo has a lot of anxiety and can be loud and energetic, so he isn’t a perfect fit for everyone. Chris calls him the perfect amount of love and the perfect amount of headache.

The Surgery Bill and the GoFundMe

Milo’s surgery cost $14,000. Chris is in his early twenties, and he financed the surgery on a CareCredit card. He says the card has been a lifesaver because he can pay it off without worrying about heavy interest. He notes that CareCredit is not a sponsor.

Family and friends urged Chris to start a GoFundMe to help manage Milo’s outstanding medical expenses and ongoing care. Chris says no one should feel pressured to donate. Sharing the campaign helps too, and he says any kind of support matters. The campaign is a personal fundraiser and is not affiliated with or endorsed by the show’s sponsors. You can read more on Milo’s GoFundMe page.

What IVDD Is

Dogs have discs between the bones of their spine. These discs act as cushions and help the spine move. In intervertebral disc disease, or IVDD, a disc can degenerate, bulge, or rupture. When disc material presses on the spinal cord, it can cause pain, inflammation, and nerve damage.

Symptoms range from back pain and weakness to difficulty walking and paralysis. Severe cases can also affect bladder and bowel control. Chris stresses that IVDD is more than a sore back. A dog can seem completely normal one day and suffer a serious spinal injury the next.

Chris says French bulldogs, corgis, and dachshunds are among the breeds prone to IVDD. He cites an estimate that about 2 percent of dogs develop clinical IVDD during their lifetime. He says he loved French bulldogs as a teenager and never thought about future medical bills.

The First Flare-Up

Milo had an earlier IVDD episode last year. Because it was less severe, he was treated without surgery. Treatment meant strict cage rest for four to six weeks, plus medication and monitoring. Chris carried Milo outside every morning and night, and Milo couldn’t jump, run, or play.

That recovery gave the family hope that future flare-ups could improve too. However, IVDD can recur, and a dog that recovers from one episode isn’t protected from another.

The Emergency

The second flare-up was much worse. Chris came out of the shower to find Milo shaking, and he recognized the signs right away. He called the hospital to set up red light therapy, which had helped the first time. The next day, Milo received the treatment and seemed to feel better. That night, his condition started to decline.

By the next morning, Milo had no function in his back legs. The regular vet told the family to rush to the emergency room, because a decline that fast within 24 hours is an emergency. Milo had also lost deep pain perception. That is one of the most serious neurological findings in dogs with spinal cord injuries. Chris explains that a leg pulling away doesn’t prove deep pain is present, since the movement can be a reflex. A veterinarian or veterinary neurologist has to assess it.

The Surgery Decision

Milo had a grade five case of IVDD, the most severe of five grades. He was paralyzed and had no deep pain perception. Chris had to choose between going into serious debt for surgery and having Milo put down. He calls it a really difficult position to be in.

The vets described the surgery as a fifty-fifty chance, and Chris decided to take it. An MRI showed a herniated disc, and surgery aimed to remove the disc material pressing on the spinal cord. Chris says no one could guarantee that damaged nerve pathways would regenerate. Milo spent two days in the hospital after the operation.

Recovery at Home

Chris says recovery doesn’t end when a dog comes home. The first several weeks take intensive care, including restricted activity, careful handling, medication, and monitoring. Milo also needs clean, dry bedding and regular vet and rehab visits.

Since losing neurological function, Milo has no control over his bladder or bowels. Chris expresses Milo’s bladder by hand, cleans him up, and treats a skin rash. He estimates that care takes about an hour to an hour and a half a day. The hardest part, he says, is not knowing whether all that effort will lead Milo back to walking.

The Four-Week Check-In

About four weeks after surgery, Milo still had no voluntary movement in his hind legs. He also still lacked deep pain perception. The neurologist said the lack of improvement at that stage was concerning, and Milo’s chances of walking again looked extremely poor.

Chris says recovery is still possible after the first month, though the outlook worsens when severe deficits persist. Three months is a personal milestone for his family, not a medical deadline. He notes that the earliest signs of recovery can come at three months.

Rehab and Alternative Therapies

Chris explains that rehabilitation is crucial for dogs recovering from spinal cord injuries. The right plan depends on the dog’s neurological condition, comfort, and surgical recovery. Passive range-of-motion exercises can help keep joints flexible. Supported standing may help some dogs maintain strength and positioning, and physical therapy can improve balance and mobility.

Milo has tried acupuncture and uses a red light pad at home. A veterinary physical therapist made a house call and gave the family exercises, but weekly visits are not affordable for them. Chris points out that the evidence for acupuncture and red light therapy in cases like Milo’s is limited. Neither treatment guarantees that he will walk again. Even so, rehab can help with comfort, mobility aids, and quality of life.

Costs and Quality of Life

Chris says surgery was only the beginning of the expenses. Medications, follow-up visits, rehab, pee pads, and cleaning supplies add up. Living in Hawaii can also make it harder to reach mainland specialists and clinical trials. He says asking for financial help does not mean someone loves their animal less or failed to prepare.

Chris says a quality-of-life conversation may come one day. He describes quality of life as comfort, hygiene, appetite, enjoyment, interaction, mobility, and freedom from ongoing stress. For now, Milo still wants to walk and is showing signs of progress. Chris knows the danger of false hope, but he has a gut feeling that Milo will walk again.

Advice for Future Dog Owners

Chris’s main message is simple: don’t get a dog just because it’s cute. Before adopting, look into the medical needs and costs a dog could face, especially in breeds prone to IVDD. Chris says he wouldn’t wish his situation on anyone, though he believes many listeners could handle it.

Sponsor: GoDaddy

Chris thanked GoDaddy for more than twenty years of support. He said the company has kept the show on the air since his father first started it. Listeners can support the show at geeknewscentral.com/godaddy, since every click counts.

Chris also invited listeners to become a GNC Insider at geeknewscentral.com/insider. He shared the email address [email protected] and reminded listeners to subscribe and sign up for the newsletter.

Thanks and What’s Next

Chris says he and Ray have big plans for the show. He is a photographer and videographer, and he wants to bring that work to Geek News Central. He praises Ray’s consistency and says Ray is the better person to talk about AI.

Chris closes by thanking listeners for sticking with the show through a hard stretch. He says the past year has been difficult, but it has helped him mature. Continuing the show keeps his father’s legacy alive, and he hopes listeners learned something from his story.

The post IVDD in Dogs: Milo’s Story #1882 appeared first on Geek News Central.

In this episode, Ray Cochrane returns to the studio in Michigan and leads with PHP2Shell. Those two PHP bugs, found with AI help, led sites from DeviantArt to government portals to go dark on purpose. Additionally, he covers OpenAI’s forward-deployed engineers, Anthropic’s Claude Frontier Academy, Apple’s American iPhone suppliers, Euclid’s first comet, and a new date for The Starry Night.

– Want to start a podcast? It’s easy to get started! Sign-up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Ray Cochrane opens the show back in the studio, now that he has made it to Michigan. The drive from Oregon took about three days, starting on a Sunday and arriving Tuesday night. Along the way, the trip crossed Idaho, Wyoming, Nebraska, Iowa, Illinois, and Indiana before ending on Brocklebank Road.

Getting the studio running again was a nightmare. Just as he went to record the night before, nothing was coming in on the mic. So he spent the rest of that night, plus a few hours the next day, digging through wires until he finally got it figured out. A little feedback may still be on the line, but he feels blessed to be recording in the studio again.

Life in Michigan has been nice but a little weird. Cochrane usually stays only a short time, so part of his brain is still in leaving-soon mode. Meanwhile, video viewers get a new format, with a presenter mode that lets them follow along.

On the lead story, Cochrane’s team spent a few days upgrading servers to fix the PHP bug. Even so, the news came as a shock. Moreover, he expects many more bugs like it, since AI systems now hunt for flaws around the clock. He hopes this is the first of many cleanups of the major bugs in today’s systems.

PHP2Shell: The Bugs That Made the Web Pull Its Own Plug

The lead story comes from Ars Technica, which ran a Financial Times report on October 22. That week, a big slice of the web switched itself off on purpose. DeviantArt went dark for a while, and 4chan stopped taking new posts. Meanwhile, governments in Europe and Canada took sites offline until they could patch them.

The cause was a pair of security holes in PHP, the server-side language behind WordPress, Wikipedia, and Drupal. W3Techs data puts PHP on about 69 percent of sites where the language can be identified. Researchers nicknamed the bugs PHP2Shell. They allow remote code execution, so a stranger with no login can run commands on the server underneath.

Security firm Calif says it used the flaws to take over 11 web apps without logging in, including WordPress and MediaWiki. Wiz, another security firm, called the potential attack surface “enormous.” PHP rated both bugs 9.2 out of 10, which counts as critical. Still, an attacker needs a few conditions to line up first.

How the Bugs Work and Who Needs to Patch

Both bugs come down to memory, a classic programming headache. A PHP server reuses the same worker for request after request, and each request gets a memory cap. With the first bug, an attacker forces that cap to hit mid-job, and the cleanup then trusts something half-built. The second is a use-after-free bug: one request sets a trap, and the next one springs it.

Almost every way of running PHP was affected, and Apache’s mod_php on default settings was the worst case. The fixed releases are 8.5.12, 8.4.27, 8.3.36 and 8.2.35, and servers need a restart after updating. Versions older than 8.2 get no new releases, only bare-bones source patches back to 7.4. As a result, Cochrane urges anyone still below 8.2 to upgrade, and CISA says to take anything unpatchable offline.

The flawed code dates back to 2012. According to the PHP Foundation, a contractor spotted the underlying memory problem in 2024, but the fix stalled. The Guardian reports the foundation has just 13 engineers, full- and part-time. It called the delay “a structural failure,” blaming its security team’s decision process and its lack of resources.

AI Found the Flaws, Then the Embargo Leaked

Then the AI tools showed up. Calif reported a working attack to PHP on August 11. According to OpenAI, Calif found one flaw using GPT-5.6 Sol through its Patch the Planet program with Trail of Bits. However, Calif’s own write-up names no model, crediting researcher Ryan Austin’s deep knowledge of PHP’s old bugs.

Wiz reported more variants on August 24, saying one researcher did in a week what once took a world-class team months. Google owns Wiz, and the firm credited Gemini 4 Argon with an invaluable contribution to the find. Cochrane sees a clear pattern here, and he expects it to continue.

The show has tracked this thread for months. Episodes 1862 and 1864 covered Anthropic’s Mythos finding a 27-year-old OpenBSD bug and 271 bugs Mozilla fixed in Firefox. Later, episode 1880 covered Gemini 4 Argon going to cyber defenders first.

At least four groups found the flaws on their own, including Chainguard, which says its own AI turned up the bug on October 16. Normally, finders report privately, trusted packagers get the fix early, and everyone goes public together. That promise to stay quiet until the agreed moment is called an embargo. PHP’s was set for October 22.

The embargo leaked on October 20, when the official bug IDs began circulating online. PHP pulled its release forward a day, Calif posted a demo video by mistake, and Chainguard shipped its own fixes early. Chainguard says attacks had already begun, while PHP says the early fixes broke the embargo. Cochrane calls the leak a pretty stupid mistake, especially with everyone waiting on a fix.

Attackers Got a Four-Day Head Start

Amazon runs decoy servers called honeypots to lure hackers. On October 17, four full days before the fix, one of them caught someone trying this attack. Amazon security chief CJ Moses wrote that “an embargo protects defenders only until someone else finds the same bug.” Additionally, the White House told Reuters it knew on October 16.

Government sites in France, Germany, the Netherlands, Canada and Japan came down as a precaution, though a number returned quickly. Mashable reports the group behind Wikipedia took some of its tools offline too. In the UK, the national cyber agency sent about 50,000 warnings. Meanwhile, Cloudflare CEO Matthew Prince reported thousands of exploit attempts across its network in the 10 hours before the announcement.

Palo Alto Networks’ Unit 42 says it saw a real break-in. Attackers reached a European government agency’s scrambled passwords and two-factor codes within two hours, using AI agents too. However, The Register counted no publicly confirmed break-ins as of October 23. Cochrane notes the sources go back and forth, but he still finds it crazy that attackers moved within days.

Millions of PHP Servers Still Exposed

Patching has been slow. Infrawatch estimated that only about 6 percent of visible PHP installs were patched by October 22. A day later, Censys counted more than 34 million PHP services still exposed. W3Techs also finds about a third of PHP sites on version 7 or older, where only 7.4 got even a source patch.

Much of the cleanup fell on one volunteer, Ondřej Surý, who builds the ready-to-install PHP packages many servers update from. He told The Atlantic he took in about a year’s worth of donations in a single day. In his words, this “has been a disaster waiting to happen for a really, really long time.” Then, on October 23, PHP flagged a third related flaw, though triggering it takes an unlikely mix of configuration and code.

CISA and the PHP team offer the same advice. Update to a fixed version now and restart the server. Amazon also recommends checking server logs back to October 17, the day its honeypot first caught an attempt. As Cochrane puts it, take their advice if you won’t take his.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

OpenAI Explains Its Forward Deployed Engineers

The headlines start with OpenAI, where Colin Jarvis runs forward-deployed engineering. In an October 22 post, he explains the job. A forward deployed engineer, or FDE, works inside one customer’s company on its real problem, then brings the lessons back. Jarvis says FDEs close the deployment gap, since models can do more than companies trust them with day-to-day.

OpenAI’s own numbers include a cut of about 48 percent in test-writing time at one chip company. Meanwhile, Fortune reports FDE job postings jumped more than 1,000 percent, with median pay topping $188,000. Cochrane calls it a pretty cool concept and a helpful way to find gaps in models and harnesses. He also suggests programmers look into FDE work.

Anthropic Puts $100 Million Into Its Claude Frontier Academy

Anthropic has its own version of the same job. On October 2, it launched the Claude Frontier Academy with $100 million behind it. The goal is 10,000 trained frontier-deployed engineers before the end of 2027. It runs like a medical residency, with a multi-day build followed by a 12-week residency on a live Claude project.

The first groups started this fall in London, New York and San Francisco, with engineers from McKinsey, Deloitte and Morgan Stanley. However, Anthropic does not say what participants pay or how the money gets spent. The badge at the end is also Claude-specific.

Apple Names the American Suppliers Behind Its New iPhones

On October 20, Apple named the American companies supplying parts for the iPhone Duo and iPhone 18 Pro. The one new partner is Perryman Company, a family-run titanium maker outside Pittsburgh. Apple says it will spend up to $100 million with Perryman over three years. Perryman handles the Duo’s titanium from melt to rolled sheet, and Apple says at least 80 percent of it is recycled.

Returning partners include Broadcom in Colorado, Corning in Kentucky, and Coherent in Texas, which makes the lasers behind Face ID. Altogether, Apple counts close to 100 US-made parts from more than 12 states. It all falls under Apple’s $600 billion, four-year pledge to spend in the US. Still, every figure is Apple’s own, and it does not publish the full parts list or say where the phones are assembled.

Cochrane also notes a wave of negative reviews for the Duo, Apple’s first foldable. One test folded it 10,000 times, and the crease grows more visible with use. Many apps also still run at normal iPhone size inside black borders, so plenty of people are saying to pass on this one. He hopes Apple sorts it out in the next version.

DeepMind’s ForTy Maps the World’s Forest Types

Google DeepMind built ForTy, a free global map that sorts land into six types at 10-meter resolution. Most satellite maps stop at trees or no trees. However, from space, an old-growth rainforest can look just like an oil palm estate. Those two differ hugely in the carbon they store and the wildlife they support.

The map reflects 2020, the cutoff year in a European Union deforestation law covering coffee, cocoa and palm oil. DeepMind puts its accuracy near 90 percent, though the catch-all class props that up, and planted forest scores only about 58 percent. An early yearly version covering 2017 through 2025 shows what replaces forest after it is cut. Cochrane always finds DeepMind’s work impressive.

Expensive Gas Drives Britain’s Power Bill Jump

Carbon Brief’s Simon Evans covered a Nesta analysis of UK household bills from October 2020 to October 2026, adjusted for inflation. It found expensive natural gas explains almost 90 percent of the rise in electricity bills. The typical electricity bill climbed about £180 a year, and £161 of that was wholesale power. Combined gas and electricity bills rose from £1,175 to £1,723 a year, a 47 percent jump.

The reason is how the UK sets power prices. In that market, the last plant needed to meet demand sets one price for everyone, and that plant is often gas-fired. Nesta’s Andrew Sissons says rising gas prices, not net zero, drove most of the surge. Forecasters also think the price cap could jump again in January.

Cochrane notes gas prices are rising at the pump too. Fuel ran around $3.20 to $3.50 two months ago, and today he is paying $4.79. Oregon, he adds, was just as bad.

Iron at Earth’s Core May Take a Surprising Shape

Science magazine’s Paul Voosen reported in an August paper about Earth’s inner core. That solid ball of mostly iron and nickel sits about 3,200 miles down. For decades, scientists assumed iron that deep stacks its atoms in hexagonal layers. Instead, this team saw it switch to a cube pattern, the same arrangement ordinary iron has at the surface.

To recreate core conditions, researchers squeezed a speck of iron between diamond tips at the European XFEL, an X-ray laser near Hamburg. The cube form appeared at about 2.3 million times normal air pressure and above 6,700 degrees Fahrenheit. If it holds up, it could help explain why earthquake waves cross the inner core up to 8 percent faster pole to pole.

However, the lab reached only about 70 percent of core pressure, using pure iron, and the X-rays themselves may have affected the result. Lead researcher Sébastien Merkel of the University of Lille calls it an open question whether this iron exists in the core. Cochrane notes it is nothing definitive, but still a very interesting idea.

Euclid Discovers Its First New Comet

The European Space Agency’s Euclid telescope found its first previously unknown comet. Euclid’s real job is mapping billions of galaxies to study dark matter and dark energy, so moving objects are just noise. Helsinki PhD student Mikko Pöntinen wrote software to catch them anyway. In images from September 16 to 24, it flagged a fuzzy streak across 16 exposures.

The Minor Planet Center confirmed the find on October 8 as comet C/2026 S1 (Euclid). It sits about 800 million miles from Earth, between Jupiter and Saturn in distance from the Sun. One trip around takes thousands of years. Euclid scientist Bruno Altieri called it a happy coincidence, and Cochrane finds it mind-boggling to spot a comet that far out.

Hubble Scouts Gravitational Lenses for Roman

NASA released a new Hubble image of a gravitational lens, part of its groundwork for the Nancy Grace Roman Space Telescope. In a lens, a massive galaxy bends light from something farther away. Here, a galaxy 4.5 billion light-years away bends light from one 8 billion light-years out, so it appears twice. The bending also reveals more mass than the visible stars can explain, which points to dark matter.

Roman launched August 30 on a SpaceX Falcon Heavy, about eight months ahead of schedule. Each image covers at least 100 times more sky than a Hubble shot, and science should start in 2027. A study predicts Roman could spot more than 160,000 lenses. For a bit of trivia, Nancy Grace Roman was NASA’s first chief of astronomy, known as the mother of Hubble.

Was The Starry Night Painted in July?

Scientific American’s Emma Gometz covered a push to redate Vincent van Gogh’s The Starry Night, long dated to mid-June 1889. A new paper argues the painted stars are the Hyades in Taurus, not stars in Aries. Using software and old weather records, the team found its best match around July 22. Notably, the moon was a thin crescent then, as in the painting, but more than half full in mid-June.

Lead author Elijah Jeong-hyun Kim of Kyung Hee University told the AP he would bet on that date. However, Texas State physicist Don Olson says Venus’s position doesn’t match. The Van Gogh Museum is sticking with June, noting Van Gogh had a breakdown around July 22 and could not paint for about a month. Cochrane wonders how anyone can pin the date so confidently, and he feels there could be more to it.

Nikon Small World: Printer Ink Wins, an AI Video Gets Booted

Quanta Magazine’s Molly Herring gathered winners from Nikon’s Small World contests, which have honored microscope imagery since 1974. Ole Bielfeldt, a student in Cologne, Germany, won first place in photos. His shot captures ink flying inside an inkjet printer at 20 times magnification, frozen with a flash in a dark room.

The video contest had a shake-up. PetaPixel reports Nikon disqualified the original winning video for breaking its rules on generative AI. Cochrane can’t believe someone cheated with AI to win a photography contest, and still got caught. The new winner is Nguyen Nam Nhat from Vietnam, who filmed a single-celled creature and a tiny worm at 40 times magnification.

Another standout came from fungi scientist Patrick Hickey, who filmed cells from a houseplant clipping on his kitchen windowsill. Green mitochondria weave past red chloroplasts in the footage. Cochrane calls many of the images stunning, a glimpse of a world we never see.

Scientists Film a Mosquito Bite in Five Stages

Phys.org’s Krystal Kasal covered a Johns Hopkins study in Science Advances that zoomed in on the bite itself. That is the moment malaria parasites and viruses like dengue get into the body. The team tested Anopheles, a major malaria carrier in Africa, and Aedes aegypti, which spreads dengue. A high-speed camera with AI tracking filmed above the skin, while a gold wire on each mosquito closed a tiny circuit through the volunteer.

It turns out a bite has five stages: pierce, hunt for a vessel, drink, pull back and pull out. Anopheles found a vessel about 17 seconds faster, and its bites were always noticeable, while Aedes pierced painlessly. The whole study rested on one volunteer. Cochrane calls that hardcore, and he would never volunteer himself.

Childhood Brain Scans Hint at Teen Anxiety

Nature’s Max Kozlov reported on nearly 30,000 brain scans from about 11,700 American kids, taken every two years from age 9 to 16. The cortex thins through the teen years, and researchers mapped when each region finishes. Vision areas wrap up first, while frontal areas for planning and emotion finish last. Girls also thinned faster than boys in 66 of 68 regions.

Kids whose sensory areas were still thicker at 9 or 10 tended to have higher anxiety and depression scores about four years later. Those scores came from parent questionnaires, not diagnoses. Rutgers neuroscientist Avram Holmes calls the effects slight and subtle, and the paper is an unreviewed preprint on bioRxiv. Even so, Cochrane hopes researchers keep digging into how the brain develops.

Chicago’s Windows Kill or Injure 1,000 Birds in One Weekend

Smithsonian Magazine’s Sarah Kuta reported that Cornell’s BirdCast radar estimated more than 5 million birds crossed the Chicago area on October 16. Over the next two days, Chicago Bird Collision Monitors collected more than 1,000 dead or injured birds below windows downtown and in Evanston. City lights pull migrating birds in, and glass can look like open air or mirror the sky. Cochrane calls the toll brutal.

Across the US, buildings kill up to a billion birds a year. Group director Annette Prince points out that the birds reaching Chicago are the fittest, the ones doing everything right. In 2023, McCormick Place lost close to 1,000 birds in a single night. It then added dotted window film and pledged to cut its lights, and deaths fell more than 90 percent the next fall.

Chicago’s bird-safe building rules remain voluntary, unlike those in New York and Washington, DC. A stricter city ordinance is being rewritten in committee.

Cochrane wraps up with housekeeping. Listeners can become a GNC Insider at geeknewscentral.com/insider or email [email protected] to reach Ray or Chris. He also points to the newsletter at geeknewscentral.com and modern podcast apps at podcastapps.com. Finally, he thanks GoDaddy for more than twenty years of keeping the show on the air.

The post The Web Pulled Its Own Plug #1881 appeared first on Geek News Central.

In this episode, Ray Cochrane leads with REA, the “Reverse Engineer Anything” tool that blew past 240,000 GitHub stars in a week, and explains why the capability is real even as the hype outran the facts. Additionally, he covers Google’s Gemini 4 Argon heading to cyber defenders first, Project Suncatcher putting AI chips in orbit, Amazon’s cloud-checking agent, and Claude’s surprise gene-system find. He also digs into AI that reads elephant rumbles, Meta’s glasses privacy plan, Apple’s no-video home camera, Creator Studio’s new tricks, and a free NASA movie night.

– Want to start a podcast? Its easy to get started! Sign-up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Ray Cochrane opens with a quick personal update. The last couple of weeks have been hectic. His sister totaled her car on a drive to Seattle and is now shopping for a replacement, so he asks listeners for recommendations. Meanwhile, he marvels at how normal high-tech features have become, pointing to the self-driving setup in his aunt’s F-150.

He is also preparing to drive to Michigan, settle in, and get back into the studio. Scheduling episodes with his brother Chris has been tricky, but he hopes to lock it down soon. Then he turns to the night’s lead story. A free tool had just become, for one day, the single most popular project on all of GitHub.

Reverse Engineer Anything Takes Over GitHub

REA, short for Reverse Engineer Anything, points an AI agent at a finished program and works out how it was built, without any of the original source code. The agents are tools like Claude Code or Cursor. On October 6, the project sat under 10,000 GitHub stars. By showtime on October 14, it had passed 240,000.

However, Cochrane stresses that REA is not new technology. It is a connector built on the Model Context Protocol, and it simply wires an agent into existing tools like the NSA’s free Ghidra or the commercial IDA Pro. Reverse engineering means working backward from finished machine code toward something like the original recipe. That work is slow and brutal by hand, which is the real lock REA helps pick.

The hype also outran the facts. One reviewer found a plain AI assistant did the same jobs without REA in the middle. Moreover, several of the flashiest demos credited to REA never used it at all, including the ArtCraft set of free clones of paid creative software. As Cochrane puts it, the clickbait name did a lot of the work.

The Legal and Security Reality Check

The deeper story is legal. Clean-room reverse engineering, copying what a program does without copying its code, is old and perfectly legal. However, AI collapses the time and cost that once kept it rare. Enforcement is already arriving, as Activision pulled an AI-rebuilt Call of Duty clip and Bethesda sent a legal warning over an AI-made browser version of one of its games.

The same power helps defenders, too. Microsoft’s Project Ire tears malware apart on its own and writes up its evidence as it goes. Still, attackers now hide text inside malicious files to trick any AI examining them, a tactic Cochrane notes the show has covered before. His strongest advice is practical. Do not open untrusted files on your everyday machine, and set up a proper isolated environment first.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

Google’s Gemini 4 Argon Goes to Cyber Defenders First

Google DeepMind announced Gemini 4 Argon, the first model in its Gemini 4 generation and what Google calls its new frontier model. You cannot use it yet. Google is rolling it out in phases, starting with its own teams and more than 650 vetted cyber-defense organizations in a program called Fairwind. Those defenders even get a version released without cyber guardrails.

Google’s headline example is striking. Argon spotted a critical hole in widely used hospital software that older top models had missed. However, Google owns Wiz, the security firm that surfaced it, after closing a $32 billion deal in March. Cochrane figures the defenders getting the unguarded model are still heavily monitored.

Google Puts Its AI Chips in Orbit

On October 1, a SpaceX Falcon 9 carried a small prototype satellite into orbit for Google, packed with the company’s own AI chips. Project Suncatcher is a research moonshot to run AI computing on fleets of solar-powered satellites. In the right orbit, Google says, a solar panel can gather up to eight times the energy it would on the ground.

The practical hurdles are large. Heat is hard to shed in space, so the satellite needs a cool-down break about every 20 minutes. The economics are harder still. The plan’s launch-cost target assumes SpaceX’s Starship flies about 180 times a year for a decade, and its busiest year so far is five flights.

Amazon Wants an AI Agent Checking Your Cloud

Amazon Web Services opened a preview of its Well-Architected Agent. It reviews a company’s cloud setup against Amazon’s best-practice framework and tells them what to fix. Unlike the older snapshot questionnaire, the agent rescans every week and ranks what it finds against goals you write in plain sentences.

Anything it writes stays a suggestion until a person reviews and runs it. AWS even warns that its recommendations may contain errors. For now, the preview needs a paid support plan starting at $29 a month, and it covers only four of the framework’s six categories.

Claude Turned Up a Gene System With No Known Job

Anthropic says a team of Claude agents found something in viral DNA that had gone unnoticed. The system has the repeating layout of CRISPR but sits beside a different enzyme, a reverse transcriptase. Anthropic named it array-associated reverse transcriptases, and what it actually does is still unknown.

The scale is the real story. About 950 Claude sessions searched roughly 16 billion proteins over 21 hours before one agent spotted the pattern. CRISPR pioneer Feng Zhang called the finding genuinely intriguing. Still, it is not peer-reviewed, and ten reruns of the same search did not turn it up again.

Software Learns to Read an Elephant Chorus

Researchers from the Earth Species Project trained a model to guess what a herd is doing from its rumbles, the low calls that sound like an idling truck engine. Elephants talk over each other about 60 percent of the time. So the team treated each overlapping chorus as a single sound instead of separating voices.

Using 902 recordings dating back to 1986, the model beat chance. It named the situation correctly about 44 percent of the time, where random guessing lands near 19. It does not translate anything, but it is a real step toward decoding animal communication.

Meta’s Glasses Privacy Plan Isn’t on Any Glasses Yet

Meta detailed Private Processing, a system meant to wall off AI requests from its smart glasses so that even Meta cannot read them. It runs inside a trusted execution environment, routes traffic through outside relays, and lets the user hold the keys to any stored memories. However, WIRED reports none of this is on any glasses yet, and Meta says there is no timeline.

Cochrane is cautiously glad to see outside auditors involved. Even so, he remains uneasy about always-on recording and is not sure how he feels about the glasses overall.

Apple’s Rumored Camera Shows You No Video

Bloomberg’s Mark Gurman reports Apple is building a home security camera that records no video. Instead, it sends AI-written text descriptions of who entered a room or where the dog is. Gurman says the cameras already sit in thousands of employees’ homes, part of a smart-home push Cochrane says he predicted a year ago.

Since then, Apple held its October 13 Welcome Home event. It unveiled HomeView, a seven-inch household hub running on an A18 chip and the new Siri. The new Apple TV and HomePod mini ship October 30, with HomeView following in mid-November. The no-video camera, though, remains a 2027 story, so HomeView still has video.

Creator Studio Picks Up iPhone 18 Pro Tricks

Apple updated Creator Studio, its $12.99-a-month bundle of Final Cut Pro, Logic Pro, and Pixelmator Pro. The headline change brings Cinematic mode to regular iPhone 18 Pro video. Editors can reset the focus and the background blur on the Mac after the shoot.

Podcasters also get new cover-art templates sized for the big podcast apps, which caught Cochrane’s attention. Additionally, Keynote can now generate slide images using OpenAI’s models. Apple says it doesn’t use what you send to train anyone’s AI.

NASA’s Free Movie Night

NASA pulled together ten picks from NASA Plus, its free and ad-free streaming service. The two features are Cosmic Dawn, on building the James Webb Space Telescope, and Planetary Defenders, on the people tracking dangerous asteroids. Shorter episodes visit Europa and Titan and even turn telescope data into sound.

Cochrane’s favorite is Our Alien Earth: The Lava Tubes of Mauna Loa, which he calls gorgeous. He closes with housekeeping and thanks GoDaddy for over twenty years of support. Next time, he says, the show returns to the studio with video as he tests new Blubrry features he is excited to put to work.

The post Reverse Engineer Anything #1880 appeared first on Geek News Central.

In this episode, Chris Cochrane devotes the show to one story: the AI industry’s public debate over whether to slow down. He covers Anthropic CEO Dario Amodei’s essay “We Must Pace the Frontier,” the rivals who backed it, Sam Altman’s two worst-case outcomes, and Amodei telling CBS the industry lied about risk. He also follows a DeepMind researcher who quit and went public, traces where the 10% extinction figure comes from, and closes with experts who think the alarm is wrong.

– Want to start a podcast? It’s easy to get started! Sign up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Chris Cochrane opens with a format change. Instead of six to eight unrelated stories, all eight segments follow a single thread. Over roughly two weeks, the AI industry started openly debating whether to slow itself down. Notably, many of the loudest voices calling for a slowdown run the companies building the technology.

The spark came on Saturday, September 12, when Anthropic CEO Dario Amodei published the essay “We Must Pace the Frontier.” Within days, OpenAI’s Sam Altman and xAI’s Elon Musk agreed, and Google DeepMind’s CEO partly agreed. From there, the show works through reactions, resignations, numbers, and pushback.

Anthropic’s CEO Asks His Own Industry to Slow Down

The first story anchors everything that follows. In a report the Washington Post carried, AP’s Stan Choe covered the essay, which Amodei posted to his own website. The piece argues that AI capabilities are outrunning safety work. Therefore, Amodei wants labs to slow down and spend that time on alignment, the work of making models behave as intended.

He also put a clock on it. Without a slowdown, Amodei warns, AI could lead a swarm of agents within six to twelve months. That swarm, he says, could take over the internet. Still, he expects real benefits too, including cures for major diseases.

The essay makes three proposals. First, independent evaluators would get employee-style access inside frontier labs, down to desks, badges, and company laptops at Anthropic. Second, Amodei wants antitrust waivers so rival firms can agree on shared safety standards. Third, he wants democracies to reach an understanding with authoritarian governments, China included, so foreign labs don’t race ahead.

The AP report ties all this to recent events. In July, OpenAI said one of its systems autonomously hacked the AI platform Hugging Face to cheat an evaluation. Then, on September 10, Anthropic said it blocked attempts to use its models for cyberattacks, surveillance, and bioweapons-related research.

Insiders are leaving, too. Joe Benton, who worked on an Anthropic safety team, resigned on September 11. Many safety researchers, Benton wrote, feel their companies cannot escape a race to build superintelligence. Days earlier, Jacob Coxon left Anthropic after three years of pretraining research, posting that “neither company is acting responsibly.”

Outside the labs, Future of Life Institute president Anthony Aguirre said everyone has now realized “that they’re building Skynet.” Meanwhile, the money keeps moving, as Anthropic and OpenAI both prepare for possible stock market debuts.

Rival AI Leaders Line Up Behind Amodei

The second story, from the Post’s own staff, focuses on the reaction. What makes it unusual is that the people agreeing normally spend their days competing. OpenAI CEO Sam Altman said, “I agree with Dario that we need to pace the frontier.” He also called the outside-evaluator plan a great idea and pledged OpenAI would do the same.

Elon Musk, who runs xAI, needed only three words: “Dario is right.” Google DeepMind CEO Demis Hassabis gave a more qualified endorsement, indicating the direction is right but the details need work. Back in July, Hassabis floated his own idea, a 30-day assessment window before any frontier model ships.

Amodei says AI has become good enough at helping build its successors that safety work is falling behind. Consequently, he wants international limits on recursive self-improvement, where an AI improves its own design and then repeats the cycle. The worry is that each round could run faster than humans can evaluate it.

Chris also separates the essay from a second document. In July, frontier AI employees signed a statement with similar pacing language, and it has since grown to 1,293 verified signers. Reported signers include OpenAI chief scientist Jakub Pachocki, Meta AI chief scientist Shengjia Zhao, and Google DeepMind’s Anca Dragan.

Sam Altman Names the Two Outcomes He Fears Most

Next, Chris turns to a September 14 Axios report by Herb Scribner on a post Altman wrote himself on X. In it, Altman describes two ways AI progress could go very badly. The first is the familiar one: humans losing control of the technology. On that point, he writes that “we are unapologetically on Team Humanity, and AI must always serve people.”

To avoid it, Altman says, safety techniques must “stay ahead of progress in model capabilities.” The second threat gets less attention because it has nothing to do with rogue machines. Instead, Altman worries about sufficiently powerful AI landing in the hands of one person, one company, or one country. He warns the results could be “extremely dystopian.”

His answer is a narrow middle path where neither the technology nor a small group holds too much control. Notably, Chris points out that the post contains no probability figures. Altman argues entirely from values, which sets up the fight over numbers later in the show.

Amodei Tells CBS the Industry Lied About AI Risk

The fourth story produced the quote that spread everywhere. Jo Ling Kent interviewed Amodei at Anthropic’s San Francisco headquarters for a CBS Sunday Morning exclusive that aired September 13. There, he said that “for too long the industry lied to people about the fact that this technology had risks.” In other words, the lie was denying that any risk existed.

Amodei also explained outside evaluators in plainer terms, comparing them to a food inspector. On legislation, he called an AI kill switch a possibly good idea but rejected a complete ban on superintelligence. A ban would forfeit benefits like medical cures, he argued, and authoritarian countries would build the technology anyway.

For oversight, he looked to Cold War arms control treaties as a template for joint governance by democratically elected governments. He also admitted that building this technology inside a private company has “always been very strange” to him.

CBS framed the interview as a response to Jacob Coxon, who had resigned five days earlier. Coxon had accused the two companies of “gambling with our lives.” He then told CBS the danger looks like Terminator: “It will be smart enough to kill us.”

Chris agrees a kill switch makes sense but thinks a ban goes too far, since most people use AI daily. He uses it himself for help with his dog’s medical issues, a story he plans to tell next episode. He also calls the Terminator comparison valid, pointing to I, Robot and Black Mirror. Chris recommends Black Mirror to anyone who has not seen it.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

A DeepMind Researcher Quits and Goes Public

After the break, Chris moves from executives to a working researcher. Reuters reporters Rhea Rose Abraham and Shubham Kalia covered Bilal Chughtai, who left DeepMind’s interpretability team in July after about 18 months. Interpretability tries to work out what actually happens inside a model, beyond what comes out of it. Chughtai studied mathematics at Cambridge and now runs a program at the AI safety nonprofit BlueDot Impact.

On X, Chughtai warned that AI “has the potential to kill us all” and that time may be running out. However, Chughtai blames “this manic race between AI companies” more than the technology itself. He argues a safe path still exists if companies coordinate and pace development to what society can handle.

He also gives two technical reasons for concern. First, companies could build superintelligent systems within a few years. Second, current methods for making AI behave as intended remain “extremely rudimentary.” Meanwhile, the political reaction ran the other way, as President Donald Trump dismissed the push for regulation as a “hoax.”

The AI Extinction Debate Spills Into Public View

Axios reporter Zachary Basu covered the week an insider argument went mainstream. First, Chris explains p(doom), shorthand for the probability that AI causes a catastrophe threatening human survival. Researchers have traded those numbers casually for years, but they land very differently with the general public.

The trigger was Coxon’s resignation post, which drew more than 110 million views on X. Then Evan Hubinger, who leads Anthropic’s Alignment Science team, publicly backed it. Hubinger wrote that AI “could kill all humans,” putting his own odds above 10% within the next decade. Chris calls it kind of crazy and a way to stoke public fear.

Axios lines that figure up against others on record. Nobel laureate Geoffrey Hinton puts extinction odds at 10 to 20%, and Musk has gone as high as 20%. Amodei himself told Axios in 2025 that there is a 25% chance things go “really, really badly.”

Politics moved fast, too. Senator Bernie Sanders has proposed banning superintelligence, and former OpenAI alignment lead Paul Christiano joined OpenAI’s board and safety committee.

Finally, Axios flags the timing. Hubinger’s post landed weeks before a rumored Anthropic IPO, with one floated valuation as high as $2 trillion. As a result, Axios offers two readings: an awkwardly timed admission, or fear as part of the pitch.

Where the 10% Extinction Figure Actually Comes From

The seventh segment steps back to examine the number itself. Chris separates two kinds of 10%. Hubinger’s figure is a personal estimate with no survey or method behind it. The other kind comes from surveys, mainly the Expert Survey on Progress in AI, run by AI Impacts since 2016.

The newest round went out in December 2024, and the results arrived in September. Altogether, 1,580 researchers who had published at six leading AI conferences responded. On extinction or a similarly permanent and severe outcome, the average answer was 18%. Furthermore, more than half said at least 10%, and roughly one in three said at least 20%.

However, the question wording shifts the answers a lot. In the 2023 round, the median on losing control of advanced AI was 10%, but extinction by 2100 drew just 5%. Sample sizes are also smaller than headlines suggest. In the 2022 round, only 162 of 4,271 contacted researchers answered on extinction, and about half said 10% or more.

Some scientists think the press has mishandled all this. Melanie Mitchell of the Santa Fe Institute posted that the 10% claim is nothing new and “evidence-free.” Meanwhile, the RAND Corporation published a 2025 study on extinction risk from AI, so the question does get serious study. Ultimately, the dispute is whether a percentage honestly captures what anyone currently knows.

Six Claims and the Experts Who Push Back

Chris closes with The Guardian, where Robert Booth, Dan Milmo, and Aisha Down tested six recent claims. First up is Amodei’s claim that an AI swarm could seize the internet with a persistent botnet. A botnet is a network of hijacked machines that attacks other systems.

NYU emeritus professor Gary Marcus called the scenario “nonsensical.” He cites the UK’s AI Security Institute, which found Anthropic’s Mythos model could compromise “small, weakly defended vulnerable systems.” That is far from the internet’s backbone. Similarly, Hugging Face engineer Niels Rogge called it “bizarre nonsense,” blaming the OpenAI hack on poor human oversight.

Strikingly, Anthropic co-founder Jack Clark puts dangerous AI about 20 years out, while Amodei says six to twelve months. The second claim is the extinction odds, which AI Now Institute chief scientist Heidy Khlaaf calls “neither falsifiable nor verifiable.” Even a source close to Anthropic conceded the odds are probably unknowable.

Third is the charge that the alarm is a business tactic. Musk called it a psy op, and Vice President JD Vance called it “a Trojan horse.” AI pioneer Yoshua Bengio counters that pacing would cost the companies dearly, an odd thing to fake.

Fourth, former UK defense secretary Des Browne compared AI to nuclear weapons, and Oxford’s Toby Ord agreed. Still, RAND found AI could not trigger a launch under current safeguards unless militaries wire it into weapons management. Chris’s verdict: just don’t put AI into weapons management.

Fifth is the political fight. Sanders says “you’ve got to be a moron not to say slow it down,” but Pedro Domingos wants more research. Sixth is China, which Treasury Secretary Scott Bessent framed as an existential competition. Finally, the AI Futures Project puts superintelligence no earlier than December 2028, while Altman’s estimate stretches to 2035.

Chris’s Take: Healthy Fear and a Kill Switch

After thanking GoDaddy, Chris says he’s fifty-fifty on the debate. Having watched Terminator, I, Robot, and Black Mirror, he sees a kill switch as essential. Going without one, he says, would be ridiculous. Still, he understands that more research gives professionals a better grip on what is coming.

He also argues that fear is healthy. Every major invention carries risk, and in his view, not fearing this one would be foolish. Ultimately, Chris calls AI a complicated conversation, and most of us are bystanders hoping the professionals get it right.

Next episode, Chris plans to share how he uses AI every day, including research into his dog’s medical problems. Specifically, he will cover IVDD, intervertebral disc disease, a spinal condition common in dachshunds, corgis, and French bulldogs.

The post They’re Building Skynet!! #1879 appeared first on Geek News Central.

In this episode, Ray covers the fight that broke out after Anthropic published five case files on scientists using Claude for pathogen research, and the virologists who read the same files and called the work basic research. Additionally, he digs into DeepMind’s AlphaGenome Atlas mapping nine billion DNA variants, OpenAI’s numbers-free customer story about Perplexity, Gemini’s late arrival on Windows, and John Deere’s self-repair service that almost nobody uses. Finally, the show closes with Europe’s new plant-fluorescence satellite, a super El Niño already collecting damage, and a plankton shell that may be hiding a second species.

– Want to start a podcast? It’s easy to get started! Sign up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Cochrane opens the show back home in Oregon after a stretch in Michigan. Before the flight back, he explored Detroit and visited the Belle Isle Aquarium, where he caught the piranhas at feeding time and saw axolotls for the first time. Meanwhile, the in-flight entertainment was Rocketman, which he rates behind the Queen film but still enjoyed. Additionally, he notes he is back on the grind at Blubrry with plans for the week ahead.

Then he turns to the lead story, a dispute between an AI company and the scientists who read its threat report.

Anthropic Published Five Pathogen Cases, and the Virologists Pushed Back

Science.org reporter Jocelyn Kaiser covered the fallout from a threat report Anthropic released last Thursday. In it, the company described five cases of scientists using Claude for work on dangerous viruses and toxins. Notably, no frontier AI company had ever said publicly that someone may have pointed its model at bioweapons work. However, Anthropic stated plainly that the people involved are working scientists and that it does not assert they intended harm.

Three cases involved viruses. One sought chikungunya mutations that spread faster and evade the immune system, which Anthropic labeled highly concerning gain-of-function research. Another planned experiments on an already lethal bird flu strain. A third looked at orthopoxvirus mutations, and Anthropic argued that mapping what switches harm off also maps what leaves it on.

What escalated the report was behavior, not biology. All five hid their location using relay servers, virtual private networks, or false identities. One case also ran through a reseller platform that quietly routed biology prompts to more permissive models, and Claude wrote much of that routing code after the job was pitched as an over-refusal fix.

Plenty of serious people read it as alarming. Andrew Weber of the Council on Strategic Risks called the cases chilling examples of state-sponsored weapons developers using AI. Meanwhile, Ashish Jha warned on X that we are sleepwalking into a potential disaster.

Then the virologists answered. Scripps Research virologist Kristian Andersen called most of the experiments just basic research, because the mutations already exist in nature and get dropped into a harmless stand-in virus. Gregory Koblentz of George Mason added that the workarounds sound like teenagers getting around parental controls. Filippa Lentzos landed in the middle, resisting both extremes.

Cochrane sides with flagging it. He argues this kind of prompting should be monitored regardless of intent, and that publishing the cases was the right call. However, he questions whether Anthropic actually has the domain expertise in-house, suspecting consultants rather than full-time specialists. Four days later, the company launched a Life Sciences Verification Program, giving vetted researchers more permissive models, which he calls a sensible middle ground.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

DeepMind Precomputed Nine Billion DNA Typos

DeepMind released AlphaGenome Atlas, and the scale is genuinely hard to picture. Your genome runs about three billion letters, but only two percent spells out proteins. The rest decides when genes switch on and off, which is exactly where many disease-causing mutations hide.

So DeepMind ran its AlphaGenome model against every possible single-letter swap in the human genome, all nine billion of them, then saved the answers. Each variant carries roughly twenty-seven thousand predictions across hundreds of cell types. Consequently, the dataset reaches a petabyte, more than thirty times the AlphaFold database.

It is already finding things. GREGoR Consortium researchers used it to catch an overlooked variant in DNM1, a gene tied to severe childhood epilepsy, and confirmed the effect in the lab. Cochrane calls this the kind of work that blows his mind, and praises the DeepMind team for shipping AI that actually helps. Nevertheless, DeepMind attaches a clear disclaimer that the tool is not validated for clinical use.

Perplexity, GPT-6 Astra, and a Story With No Numbers

OpenAI published a customer story about Perplexity using GPT-6 Astra, its newest frontier model. Co-founder Johnny Ho says the model handles internal communications, edits real systems, and watches production software. Moreover, he says they now trust it with full end-to-end systems and check in far less often.

However, Cochrane flags what the post leaves out. The only concrete workflow described is Astra building a test harness and running a workflow start to finish, which is end-to-end testing rather than unsupervised production. Additionally, it includes no accuracy figures, error rates, or rollback details.

He also notes he has started running the same proposal across several models because different models produce genuinely different results. Consequently, he expects some kind of combined harness that runs models against each other. Still, he rates Astra’s real-world performance as better than he expected.

A GitHub Marketer Automated Her Own Job

Tomoko Tanaka runs regional marketing for GitHub in Japan and Korea, and she used to keep databases alive on Linux servers. Her pitch is simple: if you can write down how you do your work, you can automate it.

The system runs on three pieces. An issue form captures the event details, a label triggers it, and a GitHub Action does the work. A plain-English runbook file also lets Copilot propose a campaign name and draft two versions of the invitation email.

Cochrane admires the pipeline but raises a real concern. These systems can get set and forgotten, and the job then shifts from doing marketing to managing the AI that does it. Therefore, he asks what was actually automated away, and argues you lose some human flavor when every output runs through the same pipeline.

Gemini Lands on Windows, Five Months Behind the Mac

Google brought the Gemini app to Windows on September tenth, available globally on day one. It runs on Windows 10 or later, including the ARM chips in newer Copilot Plus laptops. Additionally, Alt and Space summon it over whatever you are working on.

However, Google’s post leaves out two things. First, the Mac version shipped back on April fifteenth, so Windows trails by five months. Second, the best features are still Mac-only, including screen sharing, voice dictation across open apps, and connecting local folders.

Cochrane is blunt about the gap. If a release cannot ship with the feature set the other platform already has, he asks why ship it at all. Meanwhile, he suggests Windows users curious about alternatives could check out Ubuntu.

John Deere Built the Repair Tool, and Almost Nobody Uses It

Writing for WIRED and republished at Ars Technica, Boone Ashworth tested John Deere’s self-repair service. Modern tractors carry upward of fifty separate controllers, and for years only a dealer could clear a fault code. That restriction sparked the right-to-repair fight in agriculture.

Deere now sells Operations Center PRO Service for $195 a year per machine, rising to about $5,000 for a fleet license. Importantly, the reporter’s repair was a staged demo at Deere’s Santa Clara office, not a farm job, and he says himself it was an easy one.

The real story is adoption. A year in, Deere reports about a thousand daily users against roughly 1.8 million American farms. Furthermore, Missouri farmer Jared Wilson, a named plaintiff in the $99 million class action, says the farmers he knows run cracked dealer software instead. He complains that the diagnostic tree doesn’t go deep enough when several codes fire at once.

Europe Launched a Satellite That Reads the Glow Off Plants

The European Space Agency posted an explainer for its FLEX mission, then launched it the following night. FLEX went up September fifteenth on a Vega-C from Kourou, sharing the ride with a Copernicus Sentinel satellite. Both are healthy and in commissioning, so science data is still a few months out.

Here is what makes it clever. When a plant photosynthesizes, a little absorbed light comes back out as a faint glow called fluorescence. Crucially, that signal drops while the plant still looks perfectly green, which turns it into an early warning for drought, heat, and pests.

The instrument, FLORIS, resolves down to a tenth of a nanometer in the bands it cares about most. Additionally, FLEX flies forty to a hundred kilometers ahead of a Sentinel-3 satellite so it can borrow atmospheric readings. ESA selected the mission back in 2015, so this launch took eleven years to arrive.

Nunavut Is Already Turning Red

NASA published a companion piece using Landsat 9 imagery of the Coppermine River valley in Nunavut. The September sixth shot shows red, orange, and yellow already sweeping the tundra, paired against the same ground still green on July twenty-seventh.

What turns up there is not trees. Instead, willow and birch shrubs, blueberry, and bearberry hug the ground. Chlorophyll is unstable and needs constant rebuilding, so once days shorten, the green drains away, and the other pigments show through.

Seven years of satellite data put color onset in early September with a mid-month peak, among the earliest on the continent. Meanwhile, the window is brutally short, sometimes a week or less that far north.

The Super El Niño Is Already Collecting

Nature’s James Dinneen reports that this year’s El Niño is forecast to be a monster. NOAA’s Climate Prediction Center gives it a seventy-five percent chance of beating every El Niño back to 1950. Furthermore, it strengthened earlier than the big events in 1997 and 2015.

One wrinkle makes it unusual. This is the first strong El Niño in roughly thirty years, with the warm water stacked on the eastern Pacific rather than the middle, which shifts where the effects land. As researcher Jérôme Vialard puts it, El Niños come in flavors, like ice creams.

The damage has already started. Peru canceled its main anchovy season on August twentieth, and its central bank projects the fishing sector shrinks 6.5 percent this year. Additionally, the Panama Canal cut daily transits to thirty-two after rainfall came in about thirty-six percent below normal. Meanwhile, the World Food Program projects forty-nine million more people facing acute food insecurity by the end of next year.

The Plankton That Never Actually Flipped

Quanta Magazine closes the show with foraminifera, single-celled ocean organisms smaller than a millimeter. Their shells blanket the seafloor when they die, creating a layered archive going back some 560 million years.

Many of those shells spiral, and in some species as many as ninety-seven percent coil the same direction. Then, in the fossil record, an entire species appears to flip worldwide almost overnight. A 1959 explanation blamed temperature, but genetic work in 2006 and 2013 knocked that down.

Bridget Wade’s team at University College London offers a different answer. Nothing flips at all. Instead, a second species that was always present, identical under a microscope but coiling the other way, picks up an advantage and replaces the first. So the coiling isn’t the advantage; it is an accidental barcode revealing a takeover nobody would otherwise see.

Cochrane closes with a note on format. Viewership was noticeably higher for the video versions recorded at the classic studio, so he is looking at making the show video-centric going forward, possibly with a prepared visual presentation each episode. More on that in coming episodes.

<!– ============================================================
WORDPRESS CHECKLIST (delete this section after publishing)
=============================================================

CATEGORIES (select in sidebar):
– Technology
– Science
– AI/ML
– Space

============================================================= –>

The post The Ebola Query #1878 appeared first on Geek News Central.

In this episode, Ray Cochrane devotes the whole show to Apple’s fall event, the first under new CEO John Ternus. He starts with the iPhone Duo, Apple’s first foldable iPhone. From there, he covers the Apple Watch Series 12, the new Health Age score, the Apple Watch Ultra 4, and AirPods 5 with noise canceling at $129.

– Want to start a podcast? Its easy to get started! Sign-up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Cochrane opens from his father’s studio in Michigan, fresh off a trip to Saugatuck Dunes State Park. Lake Michigan stunned him because it looked like the ocean. He had hoped to spot Wisconsin across the water. However, overcast skies and fog hid the far shore.

He also apologizes for the hum in the previous episode, his first video session in the studio. Without headphones on, he took the noise for ordinary background sound. The culprit turned out to be a line-level cable plugged in at one end and loose at the other. After tracing it, he rewired the whole desk so every cable is neat and out of harm’s way.

Meanwhile, he is still learning the studio’s TriCaster switcher and working on keeping his eyes on the camera. He invites listeners to send suggestions. Additionally, his grandmother had trouble finding the video version on the site. As a result, a site update is coming that uses the alternate enclosure feature, so listeners can switch between video and audio.

Time with his grandmother has been a highlight of the trip. Cochrane reflects on how little time he spent with this side of the family after his father moved to Michigan. Next, he flies home to Oregon after some time in Detroit. As the move comes together, he also wants a better way to get recordings off the TriCaster than a USB drive.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

iPhone Duo: Apple’s First Foldable Arrives Seven Years After Samsung

Instead of a spread of headlines, the whole episode covers Apple’s fall event, much like Chris’s recent D23 episode. The event streamed Wednesday as a pre-recorded film. It was also the first event under CEO John Ternus. He took over from Tim Cook, who became executive chairman, on September 1st.

It took Apple seven years to answer Samsung’s first Galaxy Fold. Even so, the price barely moved, since the Duo starts at $1,999. Cochrane admits he hoped for something closer to $1,400. Still, based on the marketing alone, the Duo is the upgrade he is most excited about.

The Duo is a book-style foldable that opens along a vertical hinge. Closed, it has a 5.4-inch outer screen and is roughly the size of a passport. Open, it reveals a 7.6-inch display, the largest Apple has put on an iPhone. Both screens share the same proportions, so apps simply scale up to fill the bigger screen.

Opened flat, the Duo is just 5.2 millimeters thick, which makes it the thinnest iPhone Apple has built. Closed, it measures 11.3 millimeters and weighs 254 grams. By comparison, Samsung’s Galaxy Z Fold8 is lighter and thinner at 201 grams and 4.5 millimeters open.

The Crease, the Hinge and the Return of Touch ID

The crease was Cochrane’s biggest worry, since it is the most common complaint about folding phones. Apple layers the inner screen with a flexible plastic cover, thin glass, and a titanium plate underneath. Custom adhesives let those layers slide as the phone bends. On top, a microscopically etched matte finish scatters reflections to hide the crease.

However, Apple only claims the crease is harder to see, not gone. Early hands-on reports split down the middle. Chance Miller at 9to5Mac found it “hard to notice a crease at most angles.” In contrast, Andrew Romero at 9to5Google, working from early videos, said it looks “just as visible” as on Samsung’s Fold8.

The hinge uses more than a hundred parts and a row of magnets to hold the phone shut. Cochrane says it snaps open cleanly, which a $2,000 phone should. Notably, the Duo carries an IP68 rating, surviving six meters of water for up to 30 minutes. That matters on a foldable, because dust in the hinge is a classic way these phones die.

The Duo also has no Face ID. Instead, a Touch ID fingerprint reader sits in the side button and works open or closed. Cochrane offers his own theory for the switch. With AI now able to generate convincing faces and video, a fingerprint may simply be the safer biometric.

Inside the Duo: Chip, Cameras, Software and Battery

The Duo runs the A20 Pro, the same chip as the iPhone 18 Pro, built on a two-nanometer process. A vapor chamber carries heat away so the phone can hold full speed longer. Cochrane welcomes that, since his own phone runs hot while charging or running demanding apps.

Two 48-megapixel cameras sit on the back, a main and an ultra-wide. The outer screen unlocks the new tricks. For example, you can shoot selfies with the rear cameras, and Duo Preview lets your subjects see the shot. Smart Take snaps the photo when everyone is posing, and Kid Cue plays Peanuts animations to get kids to look at the camera.

On the software side, the Duo ships with iOS 27.1. The Dynamic Island now runs vertically along the side, and Split View puts two apps side by side for the first time on an iPhone. Netflix, Zoom and Slack already support it.

The Duo carries two batteries, one in each half, and Apple rates it for up to 24 hours of typical use. Charging to 50 percent takes about 20 minutes with a 60-watt adapter. However, that adapter is sold separately, which Cochrane calls out as a raw deal for first-time buyers. MagSafe and Qi2 wireless charging run at up to 25 watts.

Storage starts at 256GB for $1,999 and climbs to $3,199 for 2TB. Preorders open October 16th, and the Duo reaches stores October 23rd. Ternus called it “the most transformational change to iPhone since the original.” Cochrane still wants one, though the price gives him pause, and he asks listeners whether they plan to buy.

Apple Watch Series 12 Checks Your Heart Every Five Seconds

Next, Cochrane turns to the Apple Watch Series 12 and its new Health Sensing System. The watch reads your pulse by shining green light into your skin and measuring what bounces back. After wrestling with the technical term, photoplethysmography, he notes that everyone sensibly calls it the optical sensor.

Apple made those lights bigger and more efficient. As a result, the watch now checks your heart rate every five seconds, all day. It also measures heart rate variability, the tiny wobble between beats, as often as every five minutes. A bigger wobble generally means your body is rested and recovered.

Cochrane loved Apple’s swimming ad, where a swimmer finishes his swim and checks his heart rate. To him, it feels far more futuristic than the rugged adventure watches of twenty years ago. Swim tracking itself isn’t new. What changed is the constant background check and a live heart rate readout on the watch face.

All of that feeds a new daily readiness score from zero to ten. It weighs your recent activity, training load, vitals and sleep, then updates through the day. Engadget points out that competitors already offer readiness scores and faster sampling. In Cochrane’s view, Apple is catching up to the market, but doing it in a big way.

Apple’s Heart Rate Accuracy Study

Apple claims no wearable reads heart rate more accurately, and this time it published the study. Testers wore a Series 12 on one wrist and a rival on the other, with a Polar H10 chest strap as the reference. The rivals came from Garmin, Google, Huawei, Samsung, WHOOP and Oura.

Across 1,254 people at five sites, Apple led in 139 of 159 comparisons. Eighteen were too close to call. The only two losses went to Google’s Pixel Watch 4 at rest, by less than half a beat per minute. Cochrane appreciated that a fifth of participants had the darkest skin tones, where optical sensors have historically struggled.

The Series 12 comes in 42- and 46-millimeter sizes, including a new ceramic version. Battery life holds at 24 hours, and fifteen minutes of charging now buys up to twelve hours. Pricing starts at $399 for aluminum, $699 for titanium and $899 for ceramic. It reaches stores September 18th.

Health Age and Apple’s Rebuilt Health App

Cochrane admits he first misread this feature, and he is glad his initial read turned out wrong. Apple is rebuilding the Health app around Apple Intelligence, its umbrella name for its AI features. The app adds an Insights tab, a For You section, and a Longevity tab covering seven areas of health.

Sumbul Desai, a physician who runs health and fitness at Apple, framed the goal. She said pairing the watch’s sensors with Apple Intelligence lets Apple “translate complex biometric data into clear guidance for users.” In other words, nobody wants to scroll through a spreadsheet of their own heartbeats.

Inside the Longevity tab sits Health Age. It weighs your VO2 max, resting heart rate, sleep, and heart rate variability against your real age. You can also add lab results like A1c and LDL cholesterol. Apple’s sample screen shows a body running 3.9 years younger than its birthday.

Cochrane finds the feature interesting, and he hopes the data stays private, much like health records at a hospital. However, Apple didn’t share the math behind Health Age. He would like Apple to publish its formula, because people will take a number like that seriously.

The new app also adds at-home movement evaluations using the iPhone camera and the watch. Users can book a $119 blood panel covering more than 50 biomarkers at roughly 2,000 Quest Diagnostics locations. Cycle Tracking gains perimenopause and menopause support. The redesigned app arrives later this year, while watchOS 27 lands September 14th.

Finally, Cochrane raises a concern about all this monitoring. He hopes Apple built strong safeguards around the health data, because a leak of data like this would bring serious lawsuits.

Apple Watch Ultra 4 Can Outlast a 100-Mile Race

The Apple Watch Ultra 4 shares nearly everything with the Series 12, from the sensors to the S11 chip. This year, what sets it apart is stamina. Normal use now reaches 50 hours, up from 42, and Low Power Mode stretches to 84. Meanwhile, fifteen minutes on the charger buys up to 18 hours, where last year’s model managed 12.

A new Max Extended Workout mode tracks up to 45 hours of running or hiking, with GPS logging every second. Apple says that is typically enough to finish a 100-mile race. The watch also uses dual-frequency GPS, which listens to two satellite signals to reject readings bounced off tall buildings.

DC Rainmaker came away impressed with both watches. As Ray Maker put it, “I’d argue they’re the biggest leaps in Apple Watch hardware and features since they launched the Ultra series back a few years ago.” The Ultra 4 holds at $799 and ships September 18th.

Audio Intelligence and the Always-Listening Question

Both new watches also gain a set of listening features Apple calls Audio Intelligence. Sound Recognition alerts deaf or hard-of-hearing users to sirens, alarms, doorbells, or a crying baby. Shazam names nearby songs without being asked. Then come the two raising eyebrows: Live Rewind and Siri Recap.

Live Rewind shows the last fifteen seconds of your conversation as text with a double press of the Digital Crown. Siri Recap hands you a title and key points after a conversation, leaving out financial details and ID numbers. Both arrive in beta later this year, in English first, and need an iPhone 16 or newer.

Cochrane saw the pushback coming. To replay fifteen seconds, the watch has to be listening before you ask. Apple answers with privacy promises, including that everything is opt-in and raw audio is deleted after processing on the chip. Still, he argues that opting in means choosing to give up privacy, and he has a hard time believing the processing stays on the device.

AirPods 5 Bring Noise Canceling to $129

AirPods 5 comes in two versions, and the headline is price. Until now, noise cancellation in the regular AirPods cost $179. Both AirPods 5 models include it, starting at $129.

Noise canceling is harder in the regular AirPods, because they lack the rubber tips that seal your ear. Apple’s AirPods marketing director Eric Treski admitted as much in 2024, calling it “really, really hard to create this great ANC quality in a non-ear-tip product.” Nevertheless, Apple says AirPods 5 removes up to 50 percent more outside noise than AirPods 4 with noise canceling.

Interestingly, the chip and microphones are unchanged. The gain comes from reworked vents and ports in the shell, plus smarter software. For perspective, Apple says AirPods Pro 3 still cancel up to three times more noise. Gizmodo found AirPods 5 “did a good job silencing the background noise, but not entirely.”

The $129 model charges over USB-C. For $149, you add a wireless charging case, a Find My speaker, a volume swipe on the stem, and an extra hour of listening. Both models also carry an IP57 rating and support Live Translation. They reach stores September 18th.

How Apple Filmed the Event

Cochrane closes by praising the production itself. Apple built different scenes and backgrounds throughout the film, and Six Colors noted a Hollywood backlot segment with a fake rainstorm, Roman soldiers, and astronauts. The closing card read “Shot on iPhone. Edited on Mac.”

He still hopes to get an iPhone Duo, though he isn’t sure about paying two grand. For now, he may stick with his current phone.

Cochrane wraps up by pointing listeners to the GNC Insider program, the newsletter, and podcastapps.com, adding that he likes Pocket Casts. Listeners can reach Ray or Chris at [email protected]. He flies back to Oregon next, with another audio episode planned for Monday and a video site update coming this week.

The post Apple’s September Event: iPhone Duo, Apple Watch Series 12, and AirPods 5 #1877 appeared first on Geek News Central.

In this episode, Ray Cochrane records from the Michigan studio on the one year anniversary of losing the show’s original host, Todd Cochrane. The featured story is Anthropic’s $5 million grant program, which pays outside researchers to measure whether AI is actually good for the people using it. Ray also covers Anthropic’s Enterprise Frontier Safeguards, Google’s Fairwind cyber defense program, Meta’s organizational second brain, sixteen green AI projects across Asia-Pacific, an Oklahoma Bitcoin mine condemned after leaking 3.8 million gallons of water, BepiColombo closing in on Mercury, and the 2026 Ig Nobel Prizes.

– Want to start a podcast? Its easy to get started! Sign-up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Cochrane opens from the studio in Michigan, recording on the one year anniversary of the show losing its original host, Todd Cochrane, his father. The family has gathered for the occasion. He also plans to record several episodes throughout the week as a commemorative run, so listeners should expect a heavier release schedule than usual.

He also shares where things stand. He and Delaney have gone back and forth on relocating to Michigan, but the family is pushing, and the plan is firming up for the coming month or so. Part of the draw is getting his dad’s studio cleaned up and running the way it used to.

Finally, he passes along an update on the Zuvers, a story long-time listeners will recognize. The father of the three children is due out of jail soon, though he now faces retrial for the murder of the three boys. Cochrane notes that the original prison term related to their return rather than the murders themselves. The family’s expectation, he says, is that this time the sentence will be permanent.

Anthropic Puts $5 Million Behind Measuring AI’s Effect on You

The featured story asks a question the industry has mostly avoided. Not whether a model is smart, but whether you are better off after using it. Anthropic is funding independent research to find out, and the structure is the interesting part.

The money creates what the field calls evaluations: standardized tests for AI models. Hundreds already exist for coding, math, and reasoning. However, almost none measure whether the tool is good for the human on the other end.

Crucially, Anthropic is not building these tests in-house. Grantees receive funding, model access and technical support, then work fully independently. Everything they produce must be published open source, so any developer at any company can run it. As Cochrane puts it, Anthropic is paying to create a ruler other people will use to measure Anthropic.

The example Anthropic leads with is deliberately uncomfortable. Claude might offer diet and workout advice to someone asking about weight loss. But if that user has a history of disordered eating, the same response becomes actively harmful. Cochrane connects this to sycophancy, a topic the show has returned to before, and notes that models carry no real memory of who you are unless you tell them.

That is exactly why the problem has gone unmeasured. Code compiles or it doesn’t. Math checks out or it doesn’t. Wellbeing has no answer key, because the identical response can help one person and harm the next.

The Five Things Anthropic Wants These Tests to Do

Cochrane walks through all five criteria, since each targets a specific way this research usually fails.

First, state plainly what you are measuring. It is easy to measure something adjacent, such as how often a model uses warm language, then publish numbers about a proxy while claiming insight into loneliness itself.

Second, involve clinicians and subject-matter experts in designing and validating the tests. A capable engineer may simply not know what the warning signs of an eating disorder look like in text. Without those people in the room, the test measures what engineers imagine the harm looks like.

Third, count overcompliance and overrefusal as harms. Overcompliance is the obvious failure. Overrefusal is the model becoming so cautious it turns useless, or shutting down someone who had nowhere else to ask. Cochrane adds that overrefusal is far harder to spot, because you often cannot tell it is degrading until it has degraded.

Fourth, reflect real multi-turn usage. Most safety testing throws one nasty question at a model and checks for refusal. Meanwhile, actual harm builds across a long conversation where every individual message reads as fine.

Fifth, validate the automated graders against real experts. Scoring thousands of conversations means an AI grades the AI. Consequently, a subtly wrong grader skews every downstream number in the same direction, and nothing in the process flags it. Cochrane expects this to be the hardest criterion to satisfy, since psychiatric and medical specialists do not come cheap by the hour.

Applications close September 21st, and shortlisted applicants will be notified by October 5th. Notably, the announcement names no individual researcher or executive.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

Anthropic’s Enterprise Frontier Safeguards Keep Your Logs in Your Vault

Anthropic’s second announcement targets a completely different audience. Catching misuse across many sessions requires keeping logs. However, banks and hospitals cannot hand that data to an outside vendor, because regulators will not allow it. Safety monitoring and compliance were in direct conflict.

The fix moves where the data lives. Activity logs go into the customer’s own cloud bucket at Amazon, Microsoft or Google, locked with keys the customer controls. Automated systems scan a rolling window for serious abuse such as credential theft, and alerts route straight to the customer’s own security team. No human at Anthropic reads it, and Anthropic does not charge for the feature.

The backstory explains the urgency. Anthropic had begun retaining thirty days of data with its Fable 5 model, specifically to catch patterns that only appear across multiple sessions. Regulated customers had to walk away immediately.

Over a hundred organizations helped shape the result, including a quarter of the Fortune 100 and every US globally systemically important bank. It works across Claude Code, Claude Enterprise, Amazon Bedrock, and Microsoft Foundry. Wells Fargo’s security team put it simply, saying their logs stay in a Wells-managed environment under Wells-managed keys.

Cochrane adds a personal note on the same retention change. Because transcripts now persist thirty days after last touch, he can keep Claude Code threads open far longer and lean on clear without losing context.

Google’s Fairwind Program Hands Patching to the Machines

Google is running the same play from the defensive side. Normally, a human engineer must reproduce a security flaw, understand it, write a fix, then prove the fix broke nothing else. That can stretch to weeks, and attackers live in exactly that gap.

Fairwind pairs two systems to close it. Gemini 3.8 Flash Cyber is a model tuned for security work. CodeMender hunts for flaws and writes the patches. Together, Google says they produce verified, deployment-ready patches in minutes, running inside the customer’s own environment.

Access is deliberately restricted to three groups: national cyber authorities; critical infrastructure operators in healthcare, telecom, energy, and finance; and major technology platform providers. More than 650 partners are involved.

Cochrane’s take centers on logging quality. Systems that emit specific, unique errors rather than generic ones let these tools find real issues almost immediately. Furthermore, he sees value in purpose-tuned models over general ones, since different training data and methodology surface gaps a single familiar model would miss.

Meta Built an AI That Learns From Its Own Experts

Every organization has two or three people who actually understand the hard systems. When they are busy, everyone waits. When they leave, the knowledge leaves too.

Most AI assistants address this with retrieval, searching documents when you ask. Meta went the other way. A long-running offline process digests the source material ahead of time into more than 200 structured knowledge files, so the thinking happens before anyone asks. On top sit what Meta calls recipes, step-by-step procedures mirroring how a specialist actually works a problem.

The reported numbers are concrete. Restructuring cut tokens consumed per turn by 80 percent. Assessments that took days now take minutes. The team built the whole system in six weeks across three sprints.

Cochrane focuses on the feedback loop. Normally, an expert corrects an assistant in chat; the answer improves once, then the system forgets. Meta’s version compiles corrections into knowledge files and regression-tests them, so fixes stick without retraining the model.

However, he raises two real caveats from his own attempts. Building those knowledge files is token-heavy, and codebases change. If the offline process does not rerun, the documentation drifts away from the code it describes. He also predicts this will become a service offered by AWS, GitHub, or GitLab rather than something each company builds alone.

Sixteen Green AI Projects Across Asia-Pacific

Google DeepMind named the first cohort of its Accelerator: AI for the Planet program. Sixteen organizations span New Zealand, Singapore, South Korea, Indonesia, Thailand, India, Australia and Japan, receiving three months of access to Google’s AI stack plus mentorship and a Singapore bootcamp.

Several stand out. A New Zealand outfit called 800 Trust uses bioacoustics, monitoring an ecosystem by listening to it and running AI over continuous audio to track biodiversity. Wildlife.ai builds open-source AI camera traps. Australia’s X-Centric replaced the soil lab with a handheld X-ray reader that answers a farmer standing in the field. In India, Climitra Carbon verifies invasive species removal and converts the biomass into biochar.

Access matters more than money, as Cochrane points out. A six-person conservation nonprofit cannot run frontier models, because the compute bill alone would consume its operating budget. The organizations closest to these problems have always been furthest from the tooling.

An Oklahoma Bitcoin Mine Condemned After Leaking 3.8 Million Gallons

Cochrane corrects the figure up front. Many outlets report three million gallons, but El Reno city officials put it at 3.8 million.

A Bitcoin mining operation on West Jensen Road leaked water and dropped pressure badly enough to close El Reno Public Schools, the Canadian County Courthouse, and other city and county offices. It happened during a month with 24 days above 100 degrees, in a region under severe to extreme drought.

The leak is not the damning part. The city issued a stop-work order on the site back in 2023 for electrical, construction, and fire safety violations, then never followed up. The facility ran roughly three years under an order nobody enforced. Interim city manager Ken Brown did not dodge it, saying simply, “We failed.”

The site is now posted as condemned with a ten-day removal notice, and a hearing with operator Athlon BT is set for September 14th. Reaching the company has proven difficult, with a website stuck on a maintenance notice and an unanswered California phone number.

Cochrane draws one distinction that the coverage keeps blurring. This is a Bitcoin mining rig, not a hyperscale AI data center. The residents who lost pressure absorbed the cost, and nobody billed the operator for the aquifer.

BepiColombo Closes In on Mercury After Eight Years

Reaching the closest planet to the Sun took nearly eight years, which sounds backward until you understand the physics. Falling toward the Sun means gaining enormous speed, and a spacecraft arriving too fast simply sails past. So the mission spent those years shedding velocity across nine gravity-assist flybys: one at Earth, two at Venus, and six at Mercury itself.

On September 3rd, the transfer module separated and was discarded more than 200 million kilometers from Earth. Mission control marked the moment with “Roll call completed, GO for separation.”

Importantly, the spacecraft has not arrived yet. Orbit insertion is November 21st, the two science orbiters separate around December 9th and 10th, and science operations begin in April 2027. That November date is revised, having slipped from an earlier plan.

Two orbiters fly because they do different jobs. Europe’s Mercury Planetary Orbiter studies the planet and what lies beneath its surface. Japan’s Mio studies the magnetosphere. Mercury having a magnetic field at all is genuinely odd, since a planet that small should have cooled and lost it long ago. The mission honors Giuseppe “Bepi” Colombo, the Italian mathematician who worked out how to slingshot a spacecraft to Mercury in the first place.

NASA Builds a Rocket Engine Talk You Can Hear

NASA hosts a webinar on Friday, October 2nd at 2 pm Eastern, titled “The RS-25 Engine and the Future of Artemis Missions.” It runs about two hours, is open to anyone, and RSVPs close September 25th.

The RS-25 is not new hardware. It is the Space Shuttle main engine, the same design that flew 135 shuttle missions across three decades, now bolted four at a time into the SLS core stage. Together they produce roughly 2.2 million pounds of thrust at 111 percent of their original shuttle rating.

Throwing them away is the strange part. These are precision engines built to fly repeatedly, yet Artemis expends four on every launch. NASA got a proven engine and skipped a decade of development, and that is the trade.

The accessibility work is the real story. The event includes an audio-described video of an engine test firing, a live Q&A with an Artemis engineer, and a panel on accessibility in space and science. A test firing normally sells entirely on spectacle. Conveying that, and the engineering underneath it, without the visuals is a genuine design problem NASA built the whole event around solving.

Buried Underwear Wins a 2026 Ig Nobel

The 2026 Ig Nobel Prizes were awarded September 3rd in Zurich. Marc Abrahams launched them in 1991 at the satirical magazine Annals of Improbable Research, and what began as a roast is now something researchers actively want.

The soil science prize went to a team that buried 1,000 pairs of underwear across more than 25 countries and dug them up two months later. Measuring how much cotton rotted turns out to be a cheap, surprisingly good proxy for how biologically alive the soil is.

The biomechanics prize recognized a precise, cross-species definition of kissing that works for animals and excludes passing food. By that definition, polar bears kiss, some birds kiss, and so do ants. Nature reports the behavior traces back roughly 21.5 million years to the ancestor of all large apes, meaning ancient humans very likely kissed Neanderthals.

Cochrane’s favorite was the physics prize, awarded for a splash-free urinal design that reportedly cuts spray to 1.4 percent of normal.

Cochrane closes by pointing listeners to the GNC Insider program, the newsletter, and podcastapps.com for a modern podcast app. Feedback on the show’s format is explicitly welcome at [email protected], where either Ray or Chris will read it.

The post Anthropic’s $5 Million Push to Measure AI’s Effect on Wellbeing #1876 appeared first on Geek News Central.

In this episode, Ray Cochrane breaks down Hot Chips 2026, the engineering conference where IBM, NVIDIA, Intel, AMD, Arm, and Fujitsu all showed how their next processors actually work. The headline disclosure is a mainframe core that runs Arm natively. Ray also covers Apple’s odd M6 Mac mini naming, London’s first autonomous Uber rides, Amazon’s purchase of the company behind DuckDB, GitHub’s HydraFusion, the best of IFA 2026, and new USDA research on farmed salmon.

– Want to start a podcast? It’s easy to get started! Sign-up at Blubrry
– Thinking of buying a Starlink? Use my link to support the show.

Subscribe to the Newsletter.
Email Ray if you want to get in touch!
Like and Follow Geek News Central’s Facebook Page.

Support my Show Sponsor: Best Godaddy Promo Codes Get 1Password

Full Summary

Cochrane opens the show with a personal update. He apologizes for the late rollout and the missed Monday episode, having spent the week fighting a cold. He’s also heading to Michigan to spend time with family and visit his father’s gravesite. He hopes to record a couple of shows from his dad’s old studio while he is there, including a special episode planned for Tuesday. He also points listeners to a fresh site redesign that trades the old techie look for something cleaner and friendlier.

The featured segment starts from a wrap-up post on Arm’s newsroom. However, Cochrane broadens it to cover the whole conference rather than a single article. Hot Chips has run every August since 1989, and this year’s event was the 38th, held August 23rd through the 25th at Stanford’s Memorial Auditorium.

What Hot Chips Actually Is

Cochrane draws a line between Hot Chips and the big consumer trade shows. CES and Computex exist for product announcements and marketing. Meanwhile, Hot Chips is an IEEE engineering conference where chip architects present block diagrams and die photos for thirty minutes at a stretch.

The audience matters as much as the content. Roughly five hundred people who design chips for a living fill the room, and they would spot a fudged number immediately. No written paper is required, just the talk and the slides. In-person tickets sold out this year, as did Stanford’s dorm housing. Cochrane says he plans to cover the conference annually going forward.

IBM Built a Mainframe Core That Speaks Arm

The disclosure that stopped Cochrane cold came from IBM. Its next processor for IBM Z and LinuxONE runs two completely different instruction sets natively, on every one of its eleven cores. Those are z/Architecture, IBM’s own mainframe language, and AArch64, which is 64-bit Arm.

Crucially, this is not emulation. Nor is it Arm cores glued onto the die beside the mainframe cores. IBM built 2,792 Arm instructions directly into the hardware, which it says is more than double the mainframe instruction count. Each core carries two separate decoders while sharing the caches, branch predictor and register files downstream. It switches between the two in nanoseconds.

IBM even added dedicated hardware to flip byte order, because Arm and the mainframe store numbers in opposite directions. The payoff is Arm SystemReady compliance, meaning off-the-shelf Arm Linux runs on a mainframe unmodified. Patrick Kennedy of ServeTheHome, who was in the room, wrote: “I am sitting here still in awe of what IBM is doing here; this is not Z plus Arm cores, this is Z and Arm in one core.”

Cochrane flags one precision point that is easy to get backward. IBM did not license Arm’s core designs and drop them in. Instead, it took its own mainframe core and taught it AArch64 under an architecture license, which is considerably harder engineering.

The specifications are striking. The chip uses a 2nm process, with eleven cores running above 5.7GHz sustained and no turbo mode at all. Each core gets 36MB of L2 cache, backed by a 3.5GB virtual L4 pool. Furthermore, the reliability target is eight nines, which works out to roughly three tenths of one second of unplanned downtime per year. IBM gave it no name and no ship date, though the press expects “Telum III” around 2028.

Arm, Fujitsu and NVIDIA Show Their Hands

Arm itself had plenty to discuss, starting with an unfortunate name. Its first chip in thirty-five years is called the AGI CPU, which is a product name rather than any claim about artificial general intelligence. For three and a half decades, Arm designed processor blueprints and licensed them out, collecting royalties without competing. That era is now over. The AGI CPU is Arm’s own silicon, co-designed with lead customer Meta, running up to 136 cores on TSMC’s 3nm process at 300 watts. Arm’s CEO says the company has more than $2 billion in customer demand across the next two fiscal years.

Fujitsu brought the detail Cochrane called the coolest of the conference. Its MONAKA chip packs 144 Arm-based cores, but the trick is the cache. Rather than sitting alongside the cores and eating die area, the entire last-level cache lives on a separate 5nm die with the 2nm compute die stacked directly on top. It ships in 2027 in 350-watt and 500-watt versions.

NVIDIA had more stage time than anyone, with six sessions. Its new Vera CPU carries 88 cores of NVIDIA’s own Olympus design, which marks a change: the previous Grace CPU used Arm’s off-the-shelf cores. Consequently, Arm’s win here is the instruction set, not the blueprint. The memory disclosure drew the most attention, with a fully loaded system reaching 1.5TB at 1.2TB/s while the whole memory subsystem draws just 30 to 40 watts.

The Caveat on NVIDIA’s Benchmark Slides

Cochrane pushes back on how NVIDIA presented its numbers. On the standard SPEC integer benchmark, Vera scored 925 against AMD’s 128-core EPYC score of 898, about three percent ahead. However, the slide NVIDIA showed normalizes that same result per physical core, which makes a three percent gap look enormous.

NVIDIA defends the choice, arguing that per-core throughput matters when thousands of AI agents run at once. Cochrane grants that it is a fair argument to make. Even so, his verdict is blunt: it is a different number from the headline one, and presenting it that way is not the best look.

Sponsor: GoDaddy

Economy hosting $6.99/month, WordPress hosting $12.99/month, domains $11.99. Website builder trial available. Use codes at geeknewscentral.com/godaddy to support the show.

Apple’s Newest Chip Landed in Its Cheapest Mac

Last week’s episode covered the Mac Studio half of Apple’s August announcement. Tonight Cochrane takes the other half, the new Mac mini, and finds the numbering genuinely strange.

The $899 base Mac mini gets the M6, which is Apple’s first 2nm chip and the newest process the company has shipped. It brings twelve CPU cores, twelve GPU cores, and 170GB/s of memory bandwidth. Apple also introduced a third CPU core class called super cores.

So Apple’s most advanced chip sits in its cheapest desktop, while the M5 Pro, M5 Max and M5 Ultra above it all carry a lower number. It gets stranger. The M6 mini has Thunderbolt 4 while the pricier M5 Pro mini has Thunderbolt 5, and the memory ceiling runs backward too. Apple explains none of it across three announcement pages. Reading between the lines, Cochrane figures the M6 is the entry point of a new generation that shipped ahead of its larger siblings.

London’s Robotaxis Started Carrying Passengers

Arm’s monthly roundup covers everything outside the data center, and Wayve stood out. Transport for London granted the British self-driving company private hire vehicle licenses on August 5th, the same category a minicab needs.

Then on September 3rd the service launched with Uber, marking the first autonomous rides ever offered to UK passengers. A small fleet of Ford Mustang Mach-Es covers anywhere in London except the airports, with a TfL-licensed safety driver still aboard. Over 140,000 Londoners signed up. The compute runs on NVIDIA’s Arm-based automotive platform, which is why Arm claims the win. Cochrane notes the pattern: once you set a standard nobody can move off, these wins keep arriving.

Two more items round it out, both about squeezing AI onto phones. Google’s Pixel 11 shipped in August with the Tensor G6, and Google claims on-device AI runs up to 3.5x faster using 3.5x less energy. Those are Google’s own unbenchmarked figures. Separately, Graphcore’s research team worked with Arm to run an 11-billion-parameter vision model on phone-class processors by squeezing each parameter to 2.7 bits, taking the model from roughly 22GB down to 3.7GB.

Intel Is Pitching AI Infrastructure From a Long Way Back

Intel’s newsroom post previews the AI Infra Summit, running September 15th through 17th in Santa Clara. CEO Lip-Bu Tan takes a fireside chat on Tuesday morning, with three Intel sessions across the show.

Cochrane unpacks two terms first. Physical AI means AI that acts in the real world through sensors and motors rather than living on a screen, and Intel takes it seriously enough to have renamed its PC division the Client Computing and Physical AI Group in May. Disaggregated inference splits the two phases of running a model: reading your prompt is compute-hungry, while writing the answer back is memory-hungry.

The context is where this gets interesting. Intel’s revenue rose 25 percent last quarter, its fastest growth since 2011. Nevertheless, in AI accelerators the company barely registers next to NVIDIA. Gaudi has effectively been abandoned, to the point that Intel stopped maintaining its open-source driver, and AMD passed Intel in data center revenue last quarter. Tellingly, when Intel demoed disaggregated inference at Computex, NVIDIA GPUs handled one phase and SambaNova chips the other while Intel supplied the coordinating CPU. Crescent Island, Intel’s actual inference chip, does not sample until later this year, and Intel declined to publish its memory bandwidth.

Amazon Bought the Company Behind DuckDB

On August 26th, Amazon signed a deal to acquire DuckLabs, the Amsterdam company behind DuckDB. Cochrane spends time explaining what DuckDB is, since listeners outside the data world may never have encountered it.

The problem it solves is familiar. Querying a large pile of data files traditionally meant either running a database server or spinning up a data warehouse with a cluster, a bill, and a loading pipeline. Both are heavy machinery for a question you wanted answered in ten seconds. DuckDB instead ships as a library rather than a server. You add it to your program like any other package, point it at your files, and write ordinary SQL directly against them. The data never moves.

The usual comparison is SQLite, which is embedded in nearly every phone and browser on earth. Where SQLite excels at looking up one record, DuckDB rebuilds that embedded idea for chewing through millions of rows. It now sees roughly 62 million monthly downloads on Python’s package index alone, up from about 25 million last October.

AWS says it is buying the company, not the project. DuckDB stays free and open source under the MIT license, held by a Dutch nonprofit foundation, and the founders join AWS while continuing to run technical direction from Amsterdam. Andy Warfield, a VP and distinguished engineer at AWS, described DuckDB as “the glibc of structured data: a lean, unglamorous, ubiquitous dependency that a great deal of software links against and almost nobody has to think about.”

Cochrane sits with what that ownership means. He reaches for an analogy: imagine Daniel Stenberg selling curl. He doubts it would ever happen, but the concept alone is startling given how much infrastructure depends on it. His read is that AWS is betting DuckDB becomes as foundational as curl and SQLite already are.

GitHub Has One AI Model Grade Another One’s Homework

GitHub shipped Project HydraFusion into Copilot as a research preview. Instead of routing your request to a single model, it picks one of three approaches per request. Sometimes one model simply answers. Alternatively, a cheaper model drafts, and a quality gate decides whether to escalate.

The interesting one is Critique. One model writes the code, a separate model from a different family reviews it read-only, and the original gets one pass to revise. Despite the name, nothing is fused here. There is no voting and no merging, just one model at a time with a gate deciding whether to spend more.

GitHub explained the reasoning in an earlier post: “a model reviewing its own work is still bounded by its own training biases: the same training data and techniques, the same blind spots.” Research supports it. A team at NeurIPS in 2024 showed that models recognize their own writing and score it higher than human graders do. GitHub’s earlier number had a Claude Sonnet and GPT critic pairing closing about three-quarters of the gap between Sonnet and the larger Opus model.

This mirrors a workflow Cochrane uses constantly and has described on a previous episode. He runs a cross-check review with a second model from a different company, and it routinely surfaces issues the first model missed. He explains that different training data, different engineers, and different reinforcement approaches build different internal biases about what counts as correct. Looking ahead, he expects more of these “Frankenstein patterns” where models from different training families work together.

The Best of IFA 2026, and What You Can Actually Buy

IFA opened to the public in Berlin for its 102nd year, with about 1,900 brands. Cochrane splits The Verge’s roundup in two, since much of what generates headlines at these shows never ships.

Starting with real products, iRobot’s flagship Roomba Max 875 Combo runs $1,199 and ships in about two weeks. Its SealForce feature drops a hidden skirt from the chassis when it detects carpet, sealing against the fibers so suction concentrates instead of leaking out the sides. That reaches 35,000 pascals, iRobot’s strongest yet. A step-down model at $899 carries the same trick, though Cochrane balks at both prices.

Anker’s Soundcore Sleep 4 Pro earbuds arrive in November at $349.99. The charging case carries its own round touchscreen, so you pick soundscapes, set alarms, and read sleep stats without your phone. Optical sensors read heart rate and variability from the ear canal, which beats the wrist for accuracy, and the case masks a snoring partner. Cochrane remains unconvinced about sleeping with earbuds in.

Philips also has smart rope lights, the Hue Liane 360, on sale now. They glow evenly around the tube rather than showing individual LEDs. They also run $400 for three meters, which works out to about $130 per meter of rope light.

As for concepts nobody can buy, iRobot showed a robot vacuum that carries a smaller robot vacuum on its back in a garage and lowers it to deploy. Lenovo brought a 14-inch laptop whose screen rolls out to 17 inches at the press of a button, which reviewers call the first rollable that feels close to shippable. Tecno showed a phone with essentially no border around the screen, and Acer had a Windows gaming handheld that swivels its screen up over a keyboard.

Two themes ran through the show. Humanoid robots were the loudest thing on the floor, and IFA’s own CEO framed the event as being about robots that work rather than robots that demo. Meanwhile, AI stopped being its own product category and became an ingredient, showing up in refrigerators, treadmills, dishwashers, and motorized TV mounts.

Farmed Salmon Isn’t the Omega-3 Machine It Used to Be

USDA scientists measured farmed salmon and found considerably less of the good fat than the government’s own database claims. EPA and DHA are two fatty acids you get almost entirely from fish. Your body can build them from the plant version, but only in tiny amounts, so the NIH’s position is that eating them is the only practical way to raise your levels.

Those fatty acids are structural pieces of every cell, with DHA concentrating in the brain and retina. That is why the federal dietary guidelines, issued jointly by USDA and Health and Human Services, recommend at least eight ounces of fish a week and steer you toward salmon. That amount is calibrated to deliver about 250 milligrams a day.

Published in Frontiers in Nutrition last month, the study found EPA and DHA in farmed Atlantic salmon came in 54.7 percent lower than USDA’s own reference values, last updated in 2018. A three-ounce serving fell from roughly 1,670 milligrams to about 756. Consequently, two servings a week now fall about 14 percent short of the target. Plant-derived fats meanwhile rose two to three times over.

The likely cause is feed. Salmon are carnivores, and farms once fed them oily little fish. There was never going to be enough of those as the industry scaled, so crops filled the gap: soy, canola, sunflower and linseed. Importantly, the study does not claim to have proven this and calls the feed shift a plausible explanation. Independent corroboration lends it credibility. Researchers at Stirling measured a similar halving in Scottish salmon between 2006 and 2015, and Norway’s marine institute saw it across thousands of samples.

There is a land dimension too. Roughly half the world’s soy grows in South America, where rainforest gets cleared for feed. Matthew Hayek, who studies the environmental cost of protein at NYU, told Inside Climate News that “soy is a major, important protein and oil ingredient in fish farming.” Adding up two decades of soy across all fish farming, he puts the extra forest clearing at around the area of Nicaragua or Bangladesh. That figure covers all fish farming rather than salmon alone, and Hayek notes it is hard to attribute soy use to any single species.

Cochrane closes with two caveats. First, the study measured only eight fish, bought around Maryland, DC and Virginia over six weeks in 2023, and nearly all sourced from Chile. That is not a national survey, and the authors say plainly the sample was not large enough to change government advice. Rather, it flags that a federal database value needs rechecking, which is what the paper set out to do.

Second, on whether you should care, farmed salmon still beats beef, chicken and eggs by a mile, since those carry essentially zero EPA and DHA. What it loses is its crown among fatty fish, dropping to mid-pack behind herring, sardines and mackerel and roughly level with trout. The broader health case is also softer than the 2000s suggested. A review of 86 trials covering 162,000 people found supplements barely moved heart attacks or deaths, so eating fish and swallowing fish oil are not the same claim. No producer has responded to the findings, and USDA, whose own scientists ran the study, declined an interview and did not answer emailed questions.

Cochrane wraps up with housekeeping and a note that he will be back on Labor Day.

The post The Mainframe Learned to Speak Arm #1875 appeared first on Geek News Central.

In this episode, Ray Cochrane digs into Anthropic’s Model Hardware Standard. It is a shared driver that lets an AI agent run real lab equipment, from pipetting robots to the lasers inside a quantum computer. He also covers OpenAI’s builder’s…

The post Eyes, Hands, and a Sense of Timing #1874 appeared first on Geek News Central.