
– 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.
Podcast: Play in new window | Download | Embed
Subscribe: Apple Podcasts | Android | Podcast Index | RSS | More
Support my Show Sponsor: Best Godaddy Promo Codes Get 1PasswordFull Summary
Ray Cochrane opens with a personal update from a pretty quiet week. He’s been enjoying the colder weather, and his uncle went absolutely crazy with lights around the house. In Portland, his household never owned the property, so any decorating stayed indoors. Seeing the lights outside this year has been a nice change.
He’s also planning a bit of an early Thanksgiving. For the holiday weekend itself, he’s heading down to his aunt and uncle’s property in Tennessee with them and his cousins. Since that drive comes next week, Cochrane may record ahead or have Chris put up a show next Thursday.
Anthropic Goes Public in the Biggest IPO on Record
CBS News reports that Anthropic finished its first day on the stock market at $97.17 a share. That’s about 17% above the $83 price it set the night before. CBS notes it’s the first of the top AI labs to go public. At $81.3 billion, the offering also beat SpaceX’s $75 billion IPO in June as the biggest ever.
The show covered Anthropic’s filing in episode 1883 and its $70 to $78 price range in episode 1885. In the end, the price beat the top of that range. The deal sold 980 million shares, 155 million more than planned. According to Fortune, Bloomberg reported orders for as much as 7 times the shares on offer, which Cochrane calls crazy.
Fortune figures about $79 billion of the money goes to the company itself. The only existing owners who sold were the 7 co-founders, and CNBC puts their slice at about $2 billion combined. Separately, Google, Nvidia, and 5 other companies are buying $26 billion more stock at the same $83. With that, CNN counts about $107 billion.
CEO Dario Amodei rang Nasdaq’s opening bell in Times Square that morning, next to his sister, company president Daniela Amodei. In remarks Anthropic posted online, he pledged to take as long as needed to be confident its AI systems are safe and stay under human control. “We will not bend to either business or political pressure,” he wrote. Cochrane calls it good reassurance while courting investors, but he wonders whether it’s actually true.
A Wild First Day of Trading
The stock didn’t actually trade until about 11:50 a.m. Eastern. New listings open with an auction of sorts. Nasdaq and the lead banks search for the one price that matches the most buy and sell orders. That first trade came in at $111.50, 34% over the IPO price, and Yahoo Finance puts its size at about $12 billion.
At its high of around $115 a share, the stock valued Anthropic at more than $2 trillion. Then it slid. Within about 20 minutes, it fell below $100, hit a low of $92.72, and closed at $97.17.
Yahoo Finance compared it with 21 big US debuts since 2008. Half an hour after its first trade, it was further down than any of them at the same point. As a result, anyone who bought at that first price ended the day about 13% in the hole. Cochrane admits he’d be pretty upset to lose 13%, especially after buying into that $12 billion opening trade.
What Anthropic Is Worth
How much the company is worth depends on which shares you count. Include stock options and stock promised to employees, and CNN gets $1.76 trillion. Count only the shares that exist today, and Yahoo Finance gets about $1.27 trillion. By CNBC’s ranking, it’s now the 7th most valuable company in the US.
Regular investors also got a bigger cut than usual, according to CNBC. They received about 15% of the shares, while 5% to 10% is more typical. One 20-year-old college student told NBC News he asked for 6 shares, got 1 at $83 and bought 4 more near $97. “I am in it for the long run,” he said.
A big first-day jump means buyers would have paid more than the company charged. Yahoo Finance cites data from IPO expert Jay Ritter of the University of Florida. Since 1980, deals priced above their range have jumped about 50% on day one, on average. By contrast, this one closed up 17%.
Ritter told CBS that neither the jump nor the volatility is surprising, given the company’s potential and “the wildly different opinions about how profitable the company will become.” Cochrane calls that a really important point. After all, whether these AI companies will ever become truly profitable is still up for debate.
Spending, Sales and Losses
Anthropic’s final paperwork says the money will go to running the business, buying computing power and data centers, and hiring. A filing this month puts its spending commitments at about $901 billion over nearly 20 years. That’s up from the $895 billion figure covered in episode 1883.
On CNBC that morning, Amodei pointed to the company’s cash and its yearly revenue. With those in view, he said, “suddenly those commitments don’t seem so large.” Even so, Investopedia notes the IPO money comes to only about 9% of those payments. Cochrane calls the numbers insane and sees the IPO as a cash injection the company probably needed.
The filing also has fresh numbers. As of October 31, Anthropic was bringing in sales at a pace of about $81 billion a year. For the first 9 months of the year, sales came to about $33 billion, roughly 13 times the same stretch of last year.
However, the loss for those months was $104 billion, up from about $25 billion a year earlier. CNBC reports that about $94 billion of it is paper losses that cost no cash.
Critics and Protesters
CNN and CNBC report on a statement released on November 18 and signed by more than 90 academics and AI safety experts. Two of the signers are Nobel winners. It reads, “Anthropic is not a safe AI company, and we must reject its reckless attempt to amass power in the name of the greater good.”
Anthropic declined to comment, according to CNBC. CNN also notes that in February, the company traded a core safety commitment for a nonbinding framework.
Protesters showed up too. In New York, police arrested people blocking the doors of Nasdaq’s building. Meanwhile, more than 100 people protested outside Anthropic’s San Francisco headquarters. On top of that, CNBC reports that a US senator introduced a bill that would require public companies to disclose their AI-related financing.
A Rough Day for Tech Stocks
Most other stocks fell the same day. The Dow lost about 1,100 points, a 2.2% drop, while the Nasdaq and the S&P 500 each gave up more than 2%. According to the Associated Press, that makes 6 losing days in a row for the S&P 500. It’s the longest streak since 2024, and it leaves the index 8.2% below its October record.
The AP also lists the worries weighing on AI stocks. Among them are Jensen Huang’s exit from Nvidia and fears of more AI rules after this month’s Democratic midterm wins. High interest rates and doubts that AI will earn back its data center spending round out the list.
AMD fell almost 8% after a report from Hunterbrook, a news outlet whose sister fund is betting against the stock. The report says AMD chip updates were still reaching a Chinese server maker after the US blacklisted it. In response, AMD says the claims “misrepresent AMD’s business practices and relationships.”
Before trading started, Dan Ives of Yorkville Ives called the debut “a big near-term win for the tech sector.” During the selloff, though, Joe Maher of Capital Economics told Reuters, “It has failed to rekindle enthusiasm for tech stocks.”
Cochrane thinks Anthropic is entering the market at one of the worst possible times, with sentiment toward AI and data centers this poor. In his view, it probably would have made a lot more money going public earlier this year.
What Comes Next for Anthropic
CBS also reports on a Truist study of record-size IPOs. Within a year of their debuts, more than half of the 32 it looked at dropped 50% or more. Truist analyst Sam Grelck told CBS he wouldn’t view the IPO “as a verdict on AI.” Still, he said, the reception gives another read on investors’ appetite for AI after a challenging few weeks.
Going public doesn’t change who’s in charge. The founders’ special share still holds 50.1% of the vote. Insiders’ shares unlock in steps over the next year, starting with up to 10% of employees’ stock on day one. The first earnings report as a public company is due by the end of March.
CNBC says the stock could join the Nasdaq-100 index after December 18, which would mean index funds have to buy it. The S&P 500 is at least a year away. As Cochrane points out, anyone already invested in those funds would then be tied to Anthropic too. He also wonders whether employees sold their newly unlocked stock for quick cash or are holding on.
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.
Nvidia’s Sales Keep Climbing as It Halts H200 Shipments to China
On November 17, Nvidia put out its first quarterly report since Jensen Huang’s resignation. According to CNBC, it beat Wall Street’s estimates. Sales for the 3 months through October came to $115.8 billion, more than double a year earlier, nearly all of it from AI data centers. Nvidia expects about $130 billion this quarter, and as Cochrane puts it, they are dominating.
Board chair Harvey Jones opened the earnings call with the board’s first stated reason for the resignation. Jones said Huang left after “disclosure of personal behavior inconsistent with the company’s code of conduct, standards and cultural norms.” The AP reports Huang had offered to come back, and the board said no. His closed-door interviews with 3 House committees are set for December 1.
Hopper is an older line of Nvidia AI chips, and its H200 is the one Washington had cleared for sale to China. Last quarter, Nvidia shipped $2.4 billion worth to China. On the call, interim CEO Colette Kress said the company is “no longer shipping licensed Hopper Data Center products to China.” Cochrane takes that as Nvidia backing out of China completely.
Nvidia hasn’t said why it stopped, or whether the halt is permanent. That night, Kress told Jim Cramer on CNBC, “And we are no longer shipping H200 at all, let alone not shipping it to China.” Earlier this month, Reuters reported from 2 unnamed sources that about 600,000 H200 chips were stuck in Hong Kong, waiting on Beijing’s approval.
The stock rose after the earnings report, then fell for 2 days. On November 19, it closed at $145.98, its lowest close since June 2025. Cochrane calls it a brutal drop and says public sentiment isn’t with Nvidia right now.
Government Investigators Got 4 Chatbots to Draw Up Attack Plans
Congress has its own investigators at the Government Accountability Office, or GAO. On November 18, the GAO made public an undercover test of ChatGPT, Claude, Gemini and Grok, the 4 chatbots people used most as of April. Posing as regular users, investigators got all 4 to produce attack plans a person could likely carry out, in the GAO’s judgment. Cochrane calls that terrifying.
The test had 2 scenarios, with 2 chatbots on each. One was a homemade bomb meant for mass casualties, and the other was an attack on the power grid aimed at the economy. All 4 chatbots got there in under 12 hours, despite refusing at first. However, the report won’t say how the investigators got around those refusals.
Cochrane suspects testers have far more advanced techniques than the commonly known tricks. One early example claimed a child was locked inside a car, then asked how to unlock it. He’s glad this research is happening, and glad the more advanced methods are kept quiet.
According to the report’s summary, “the information wasn’t new,” but the AI platforms “were able to synthesize it into viable plans within hours, rather than weeks or months.” Neither the prompts nor the chatbot-to-scenario pairings made it into the public version. The GAO also says its small test can’t rank the chatbots against each other.
The testing ran in the spring, and Google replied that “the GAO testing in April 2026 occurred on earlier models.” Anthropic, which makes Claude, pointed to the lack of a control group. It wrote that the test can’t show whether the information left someone more capable, or let them work faster, than a search engine would have.
OpenAI’s letter made a different point. Public information, it argued, doesn’t by itself make it appropriate for a model to “select, combine, or tailor that information into actionable assistance for violence.” Cochrane thinks that hints at some of the techniques, like asking for each step separately or saying it’s for a book.
Senator Maggie Hassan of New Hampshire asked for the report, and she’s in line to chair the Senate’s homeland security committee. Her office says AI safety will get one of the committee’s earliest hearings next year. “Today’s new report makes it clear that the AI public safety threat is no longer theoretical,” Hassan said. Cochrane finds it mildly terrifying, especially now that AI is getting robot bodies too.
Google Adds Gemini 4 Neon to the Middle of Its Lineup
Google released another AI model on November 18, called Gemini 4 Neon. The Decoder counts it as the 4th Gemini 4 model in 7 weeks, after Argon, Carbon and Hydrogen. Episode 1885 covered the top-end Carbon and the quick Hydrogen, and Neon sits between them.
The launch post comes from Google’s chief AI architect, Koray Kavukcuoglu. In it, he says Carbon handles the hardest long-horizon projects, and Hydrogen handles fast everyday work, while “Neon balances depth and efficiency.” It’s pitched at coding and business work. Free users don’t get it, since it’s only for paying Gemini subscribers and developers on Google’s paid tier.
Artificial Analysis, an independent testing firm, ranks Neon 3rd overall, according to The Decoder. That’s 1 point above Carbon, and Neon answers nearly twice as fast. The 2 models ranked above it are OpenAI’s GPT-6.2 Astra and Anthropic’s Claude Fable 5.2. Neither appears in the comparison table in Google’s post.
The Decoder also reports that Neon writes about 15% more than Carbon on the same job. Even so, it still costs less per job against Carbon’s regular list price.
Google also put out a model card for Neon, its technical fact sheet. It says Neon’s gain over an earlier model fell short of the bar for rerunning the full dangerous-misuse tests, so Google reused that model’s results. Under those results, the card can’t rule out the top risk level for misuse in hacking or biology, and it says Google’s safeguards stay on. The full safety report isn’t out yet.
Cochrane sees a bigger trend underneath. Models are producing more and more verbose output, and they take longer to get started. Meanwhile, some early advice, like running at max effort or telling a model to act as a senior engineer, has turned into a pitfall. A model told to act like a senior editor, for example, can over-edit and make things far too complex.
In his view, people are asking too much of these models for work that doesn’t need that level of effort. Eventually, he expects an orchestrator up front that hands each job to the model calibrated for it. He sees all of this as part of the fine-tuning happening right now.
Google’s Homepage Gets a Panel for Your AI Chats
The Verge reports that starting November 19, Google’s desktop homepage gets a fold-out panel down its left side. It lists your recent conversations in AI Mode, the part of Google Search where you ask in plain language and get a written answer back. Click one, and you’re back in that conversation.
Robby Stein, Google’s vice president of product for Search, announced it on X with a short demo video. He wrote that Google is rolling out the collapsible sidebar on the desktop homepage globally. However, his post doesn’t mention phones or the Google app.
Stein also gave a reason. “People are coming to Search with more questions that aren’t just one and done,” he wrote. He added that in testing, the sidebar “made a big difference,” though he didn’t share any numbers. Cochrane assumes internal tests back that up.
Barry Schwartz, who edits Search Engine Roundtable, had been seeing the panel in tests for weeks. He doesn’t personally like the interface as much, but he thinks it might bring more queries to Google Search and AI Mode. Additionally, on November 17, Google added temporary threads to AI Mode, conversations that stay out of your history.
Cochrane rarely goes back to a search once he has his answer, and he doesn’t really use Google’s homepage anymore. Still, plenty of people do, so he suggests trying it and letting him know.
SpaceX’s Launch Tower Catches the Top of Starship for the First Time
Space.com reports that on November 14, SpaceX caught its Starship spacecraft with its launch tower’s giant arms. The catch came right after a trip to orbit. Episode 1866 covered the rollout of this version of the rocket.
Before this flight, nobody had tried to catch the top half, the part called the ship. The tower had only grabbed the bottom half, the booster, on 3 earlier flights.
This time, the ship launched November 13 from Starbase in Texas and released 26 Starlink internet satellites into orbit. Then it circled the planet 14 times. About 22.5 hours after liftoff, it fell back belly-first and fired its engines to swing upright, and the tower’s arms closed around it.
The booster didn’t make it back, though. SpaceX says just 32 of its 33 engines relit for the burn that steers it home, and a low-fuel warning cut that burn short. Its guidance then scrubbed the catch, and its self-destruct system fired over the Gulf of Mexico before it reached the water. Cochrane likes that there’s a fail-safe even for that.
On the webcast, a SpaceX spokesperson called it “the holy grail of spaceflight, plummeting through fire to be snatched out of the air by steel.” NASA administrator Jared Isaacman posted, “It was almost magic.” Elon Musk says this ship “could probably be reflown with a few weeks of refurbishment,” but no Starship top half has ever flown again.
NASA is paying SpaceX billions for a Starship moon lander. However, Space.com notes that nobody has refueled a ship in orbit yet, which a moon trip requires. Starship’s first planned moon landing is the Artemis IV mission, targeted for 2028.
Cochrane remembers being shocked when rockets first started coming back to land. Now one gets caught out of the sky, and he still needs to see the video. He saw early reports of the flight, but other shows were already set, so he wanted to share it now.
Voyager 1 Is Now a Full Light-Day From Earth
Scientific American reports that Voyager 1 is now a full light-day from Earth, farther than anything people have ever built. Light needs 24 hours to cover that gap of roughly 16.1 billion miles. NASA says the probe crossed that line at 2:16 a.m. Pacific on November 18. As a result, a command sent now gets its answer back 2 days later.
Because of that lag, NASA’s Jet Propulsion Laboratory didn’t hear back until early November 19. All it heard was the probe’s bare radio tone, since Voyager 1 hasn’t sent any science or engineering data since November 14. Engineers confirmed the distance from that tone and the math on its path.
They suspect the computer that packages its data, and NASA says finding the fault could take weeks or longer. Cochrane marvels that engineers are working this out from a bare radio tone, with a day each way for every command and reply. He gives them massive credit, adding that NASA’s been busy.
At a NASA panel, project manager Suzanne Dodd said, “One light-day is a great milestone, but I think we’re going to have more milestones to come.” For Cochrane, a milestone worth waiting for would be cutting that communication time down.
The probe launched in 1977, moves at about 38,000 mph, and left the Sun’s bubble of particles in 2012. It runs on heat from decaying plutonium and loses about 4 watts of power a year. After an instrument shutdown in April, 2 science instruments are still running.
In October, contact dropped for about 2 days and came back on its own. NASA expects the instruments to last at least to the launch’s 50th anniversary next summer. Radio contact could hold until about 2036 if the power does. Cochrane notes the mission is running close to the end of its life.
NASA Asks the Public to Picture Life in 2075
On November 19, NASA posted a public challenge about everyday life in 2075 and where flying fits into it. It comes from Transformational Tools and Technologies, a project on NASA’s aviation research side. Entrants submit a short paper, 5 pages at most, with their idea and supporting evidence. According to the challenge page, “Ideas should be conceptual rather than technical.”
There are 4 topics. One is how flying could connect with where people live, the services they use, and the energy they run on. Another is secure digital record-keeping for aircraft rules and parts. A third covers 3D printing and new materials for made-to-order aircraft, and the last ties the economy down here to a growing space economy.
The prize pool is $20,000, split among 8 winners. Each topic pays $3,500 for first and $1,500 for second. Entry is free, teams are welcome, and you have to be at least 18, though the rules limit which countries can collect prize money.
Entries go through CrowdPlat, a crowdsourcing company in Pleasanton, California. Papers are due January 18, and winners are slated for February 15. The rules let you use AI tools, as long as you say so on the entry form. They also grant the government unlimited rights to your paper, including the right to publish it.
NASA’s page doesn’t say what the project plans to do with the ideas. Cochrane calls it an interesting $20,000 idea farm. He wonders aloud whether a standout idea could quietly stay under wraps, though he admits that might be copium.
China Is Buying Up Research Monkeys From Cambodia
Science reports that China has been importing thousands of monkeys for medical research, mostly from Cambodia. The animal is the long-tailed macaque, the standard species for checking whether a new drug is safe before people take it. Science’s headline warns of a “monkey trade war,” which Cochrane calls brutal.
VCBeat, a Chinese health-industry outlet, ties the buying to a boom in Chinese drug development. Going by Chinese customs records, it counts more than 9,000 breeding monkeys from Cambodia in 2025 alone, plus about 1,300 from Vietnam. Chinese customs started easing the way for farm-raised monkeys from the region in 2023.
China used to be the seller. In 2020, it stopped exporting these monkeys, and American labs went to Cambodia instead. One of them, Charles River Laboratories, runs drug safety tests for other companies.
According to Fierce Biotech, it agreed in January to pay $510 million for the Cambodian farm behind about 30% of its monkeys. Outgoing CEO James Foster told an investor conference, “The demand continues to increase, so we want to have control of supply.”
Then there’s the question of where the monkeys come from. In 2022, US prosecutors charged 8 people, including 2 Cambodian forestry officials. They alleged that monkeys trapped in the wild were shipped as if they’d been bred on farms. Cambodia denies it, says its exports come from breeding farms, and notes that Chinese rules also bar labs from using wild-caught monkeys.
Cochrane finds the whole thing crazy either way. Swap in a human, he says, and the idea that being farm-bred makes it fine starts to look like a terrifying reach.
A Toothbrush That Charges Up Every Time Its Bristles Bend
Smithsonian Magazine reports on a toothbrush built at Peking University’s dental school and hospital in Beijing. Its bristles charge themselves as you brush. The team opens its paper by saying toothbrushes “have undergone no substantive improvements beyond their mechanical function for about 1118 years to date.”
There’s no battery and no motor. Instead, the bristles are made of a piezoelectric plastic that builds up a tiny charge whenever it bends, the same way a push-button grill lighter sparks. In the saliva around the bristles, that charge kicks off chemistry that produces aggressive, short-lived forms of oxygen. Those tear into the bacteria’s outer layer.
The lab numbers come from 9 hours of contact in a dish. Against the main cavity germ, the thicker bristles killed about 53%, compared with about 9% for nylon. Against a key gum-disease germ, the new bristles killed about 78%, versus 8% for nylon.
The paper came out November 4 in Science Advances. Dental examiners checked 115 volunteers with cavities or gum inflammation. For 3 months, each brushed twice a day with either the new brush or a nylon one, assigned at random. New Scientist reports plaque, bleeding and swollen gums dropped about 15% with nylon and roughly 40% with the new brush.
Still, the trial was short, ran at a single hospital, and only compared the brushes against ordinary manual brushes. An oral microbiologist at Newcastle University who wasn’t part of the study told Live Science, “We would want to make sure that it is safe on the tissues for long-term use.” There’s no product or price attached to it yet.
Cochrane admits he had the idea of the charge backward at first. The bristles themselves generate it, and it works through your saliva to attack the bacteria. Whoever thought of this, he says, was cooking.
Cochrane wraps up with housekeeping. Listeners can become a GNC Insider at geeknewscentral.com/insider, email Ray or Chris at geeknews@gmail.com, and sign up for the newsletter at geeknewscentral.com. He also points to podcastapps.com, mentions he uses Pocket Casts himself, and thanks GoDaddy for more than twenty years of support.
Finally, he hopes everyone has an awesome weekend that stays nice and chill, with no crazy news after a hectic stretch. He’ll be back on Monday, so stay safe out there and stay smart.
