OpenAI Banks $11 Billion
OpenAI has received $11 billion toward a funding round that could reach $20 billion at a $1.25 trillion valuation, Bloomberg reports. The initial investors have already wired their money, with up to $9 billion more expected in the coming weeks. The cash would help finance the company’s infrastructure commitments ahead of a planned 2027 IPO.
Nvidia Board Probes Huang
Nvidia’s board opened an independent investigation after reports that CEO Jensen Huang had been targeted by Chinese intelligence. In a memo to employees, Huang admitted an affair, apologized to his family and staff, and said he would cooperate fully with the inquiry. The investigation will examine the national-security concerns raised by the reports.
Claude Helps Prove Quantum Conjecture
Physical Superintelligence says its research system, running a non-public Anthropic model, found a proof of the Aaronson-Ambainis conjecture during a three-day run. Lean checked the proof, and the team published the manuscript and code. The result limits quantum speedups for certain yes-or-no problems on random inputs.
Thinking Machines Nears $5B Round
Mira Murati’s Thinking Machines Lab is nearing a $5 billion funding round led by Accel at a $40 billion pre-money valuation, Business Insider reports. The proposed financing would follow last year’s $2 billion seed round. The startup has launched Tinker for model fine-tuning while competing with Meta and OpenAI for researchers.
Apple Pushes Agentic Siri
Apple CEO John Ternus is pushing AI chief Craig Federighi to accelerate Siri, Bloomberg’s Mark Gurman reports. Finer control of Apple’s own apps, such as editing photos and organizing Safari tabs, is expected with iOS 27.3, followed by third-party app actions in the spring. More persistent, goal-driven capabilities could follow as early as next year.
Meta Strategy Chief Joins Fireworks
Dave Wehner is leaving Meta to become CFO of Fireworks, an AI inference and model-training platform valued at $17.5 billion in its July funding round. Wehner served as Meta’s CFO from 2014 to 2022 before becoming strategy chief. Fireworks CEO Lin Qiao says he will help scale the startup’s finances and operations.
Karpathy Builds TristanTube in Minutes
Andrej Karpathy says he built TristanTube with Claude Code in three minutes, turning Tristan Hume’s YouTube channel list into a collection of about 47,000 videos with a TV mode. It follows HackerTube, his earlier project that used Claude to sift 43 million Hacker News comments into 6,454 recommended videos.
Tesla Liable for Solar Fraud
A San Diego judge treated Tesla as liable on fraud claims as a sanction for withholding evidence in a Solar Roof lawsuit. Tesla had already conceded breach of contract and negligence. The company is seeking appellate review, arguing the sanction was excessive. The order concerns one homeowner’s case, with damages still to be determined.
Harbinger Raises $240 Million
Electric truck maker Harbinger raised $240 million in Series D funding co-led by Tiger Global, Capricorn and Ridgeline. The Garden Grove, California company says the money will expand manufacturing and support energy storage, power generation, autonomy and defense applications. The financing follows a FedEx order for 2,000 trucks valued at more than $300 million.
Amazon Reboots Home Robots
Amazon’s Personal Robotics Group is hiring for a new home robot, CNBC reports. The effort includes Fauna Robotics, which Amazon acquired in March, and its Sprout humanoid. Job listings describe autonomous operation around family members. Amazon says it continues to support Astro and is applying lessons, talent and experience from that project to its other robotics efforts.
FDA Probes Zolgensma Cancer Risk
The FDA is investigating tumor reports following treatment with Novartis’ Zolgensma gene therapy for spinal muscular atrophy. In two cases, molecular analysis found the therapy’s vector DNA had integrated into a cancer-related gene region. The agency is reviewing labeling and long-term safety studies, while saying the treatment’s benefits continue to outweigh its risks for approved uses.
Jensen Huang admits affair as Nvidia probes spy claims
Nvidia’s board of directors just ordered an independent review of reports that CEO Jensen Huang was targeted by a “foreign intelligence effort,” with Huang admitting to an extramarital affair in a letter to staff and saying, “There is no excuse for what I did. I am deeply sorry.”
- Sec. of State Marco Rubio reportedly oversaw a spring memo on Jensen and Alice Huang, an Nvidia customer exec it linked to China’s top spy agency.
- After calling their meetings “completely normal” last month, Jensen told staff, “I was unfaithful to my wife” and reportedly confirmed the relationship to the board.
- Reporting has yet to establish what, if anything, Ms. Huang (aka Huang Le, no relation to Jensen) learned from him, while China denies the spying claims.
- Outside lawyers and independent directors will run the review as Jensen stays CEO, with the board seeing “no reason to believe” any secrets were compromised.
- Jensen also has until Nov. 11 to answer a House GOP request for a private briefing, with more congressional investigations expected over the relationship.
AI stocks are already reeling from the midterm election results. Now, a spy scandal around the CEO of one of the industry’s biggest companies is unlikely to help either. Nvidia touches nearly every aspect of the AI industry, and Huang is the frontman of it all, so all eyes will be on what the board finds and his future with Nvidia.
Turn your X bookmarks into a personal research desk
X moves quickly, and saving posts is easy. Finding the useful ideas again is harder. In this guide, you’ll connect Grok Bot to X, review a week of bookmarks, spot repeated themes, and turn one finding into a post draft.
Step-by-step:
- Check the difference between free Grok chat and Grok Bot. Bot is included with Premium, Super, and Ultra on the X Subscription. X Chat is still rolling out
- Connect the official X plugin with the Grok Bot. Check that it is linked to the same account where you save posts
- Bookmark 8 to 12 useful posts about one topic. Choose examples, questions, and problems, not just launch announcements
- Message your Bot in X Chat and ask it to group the bookmarks into themes. Check the direct links and compare each claim with the original post
- Choose one supported theme and ask for an X post draft. Review the wording and sources yourself; leave publishing off
Pro tip: A set of bookmarks can show what you have been reading, but it can’t prove what all of X thinks. Keep the source links with every theme.
Anthropic’s Claude ‘dreams’ to fix its own memory
Anthropic just published its own take on AI “dreaming,” sharing how Claude agents use downtime to fix and reorganize their memory, and what it’s learned about what an AI should actually bother remembering.
- Memory refers to notes Claude saves as files and reads back in later sessions, so it can build on past work instead of starting from scratch each time.
- Unlike ChatGPT’s version, which previously built a profile of the user, Anthropic’s dreaming has Claude check its notes against logs of past work to catch errors.
- Anthropic said newer models have gotten much better at judging what to save, and that explaining to the AI why memory matters beats rules on what to write.
- In tests on Pokémon, Fable 5 earned four badges with only 10 memory files, while an older Sonnet 3.7 got just one using 413, and Opus 5.5 beat the game entirely.
It feels like not too long ago that the early iterations of ‘memory’ in AI assistants were a comically simple scratchpad with bare-bones notes on the user. Now they can recall past chats in detail down to the most obscure reference, and the next step is agents that continuously remember the job and improve at it the longer they work.
AI Wants Your Work Files. Who Gets Paid?
Your inbox finally has a buyer. Somehow, you still don’t get a raise.
They’re coming for your 2 a.m. panic Slack threads. Brokers are dangling up to $2 million for corporate archives to train the next wave of AI models, as developers grow so starved for data they’ve already resorted to buying and shredding rare books.
Turns out, AI needs to learn how humans actually work—meaning how a simple request devolves into seven passive-aggressive revisions. The CEO of AI data broker Mercor notes that simulated data just can’t replicate the beautiful disaster of real corporate workflows. Your office chaos is officially a premium feature.
While the C-suite can turn your digital exhaust into a quick payday, everyday workers rarely see a dime or even a heads-up. Unlike data-exchange startup Verb’s consumer marketplace, where you actually get to opt in to selling your data, the corporate fire sale happens entirely above your pay grade.
Brokers like Mercor pinky-swear they scrub identifying details, but privacy experts aren’t buying it. A nameless chat about a root canal or custody battle is still pretty identifiable. Plus, lawyer Daniel Goldberg warns that linking these records can easily unmask individuals—and deleting a name doesn’t magically make your company’s secret sauce safe for public consumption.
Despite the gold rush, pesky things like NDAs, privacy laws, and common sense still apply. And before you get any bright ideas about a side hustle, remember you probably don’t own the rights to sell your own work files either.
Your hard-earned expertise is being used to train systems to automate your job. A quick payout for yesterday’s files might give an AI permanent access to your company’s institutional knowledge—and burning through employee and customer trust isn’t something you can fix with a software patch.
Claude Gets Bank Access Without Spending Power
Finally, someone to interrogate your subscriptions who can’t start another one.
Anthropic has launched Personal Finance in Claude—first spotted in testing as “Claude Money”—inviting you to connect your bank and credit card accounts through Plaid for a read-only roast of your spending habits. The US beta is rolling out to Pro, Max, and Ultra subscribers. Thankfully, it lacks the ability to actually transfer, manage, or YOLO-invest your remaining funds.
Claude can now spot duplicate charges, explain exactly why your utility bill looks like a ransom note, and chart spending across accounts. It views current balances, credit limits, and transaction histories going back 24 months. Because authentication happens directly via Plaid, the chatbot never intercepts your actual banking passwords.
That boundary matters as increasingly nervous banks restrict credential-sharing agents, warn about automated-purchase fraud, and face regulatory grilling over AI controls. Read-only access limits the damage Claude can do, though it still exposes a highly intimate, tragic record of your late-night impulse buys.
Anthropic pinky-swears that chats within its dedicated finance hub are excluded from model training, though standard privacy settings still govern any financial panic you share in general conversations.
OpenAI’s similar Plaid-connected tool began reaching US Free and Go users on Oct. 2, but ChatGPT’s finance discussions gleefully adhere to your universal model-training preferences. Claude’s hub is also blissfully free of advertisements and sponsored loan pitches.
To try Claude’s version, open Personal Finance in the sidebar and choose which accounts to share. Just remember to check important figures against your actual bank, as balances can lag. Unlinking stops future access, but existing chats, shared links, and financial artifacts require separate, manual scrubbing.
Your bank balance has an AI interpreter. Your impulse purchases have a witness.
SpaceX Recruits Qualcomm for Its 5G Push
SpaceX has tapped Qualcomm to provide the underlying hardware for Starlink Mobile’s planned nationwide US 5G network. The massive rollout is slated to start in 2027 and ultimately encompass tens of thousands of cell sites under a new multiyear agreement.
Qualcomm’s Dragonwing processors and baseband software will power the nodes, utilizing Massive MIMO arrays to boost capacity and signal reach. SpaceX still has to actually assemble the equipment and build the rest. Sadly, a “carrier in a box” isn’t available on Amazon Prime just yet.
The endgame is a combined satellite-and-cellular competitor to AT&T, Verizon, and T-Mobile. Orbital satellites can reach remote outdoor areas, while ground infrastructure supplies capacity and indoor coverage. To help push those signals through walls, SpaceX has also agreed to buy an 800 MHz spectrum portfolio, pending FCC approval.
Rumored Boost Mobile talks could provide a shortcut to an established subscriber base. But no deal is confirmed, and since Boost just rents its capacity from AT&T and T-Mobile—the exact gatekeepers who already rejected Musk’s network advances—an acquisition wouldn’t magically deliver an independent physical network.
For consumers, success could mean more coverage options and pressure on carrier pricing. Businesses could gain another option for field teams and remote operations but should verify coverage, support, and compatibility before switching. Relying on a single vendor for both satellite and cellular access also eliminates true path diversity for failover. A 2027 construction start leaves plenty of work before those benefits arrive.
SpaceX can land rockets on floating barges. Landing your enterprise phone contract, however, still requires actual customer support—a frontier Elon hasn’t quite conquered yet.