OpenAI claimed three fresh breakthroughs in math, physics, and computer science — including that the Exponential Time Hypothesis is false. Apple unveiled the M6 MacBook Pro. Microsoft offered buyouts to ~60K workers, a quarter of headcount. Plus, Anthropic S-1 nuggets.

• OpenAI announces three breakthroughs in computer science, math, and physics, including that the Exponential Time Hypothesis, a longstanding conjecture, is false (OpenAI)
• OpenAI used external “shepherd” experts to guide its computer science and physics results, while its AGMAI panel of math advisers recommends AI companies drop the practice, citing unequal access, borrowed reputation, and biased selection (The Verge)
• Apple unveils the 14-inch MacBook Pro with M6, with a 12-core CPU and a 12-core GPU, up to 32GB of unified memory, and Wi-Fi 7, from $1,999 on November 6 (Apple)
• The M6 MacBook Pro, the new base model replacing the M5, charges to 50% in 30 minutes, adds Bluetooth 6, and opens for pre-order tomorrow, October 28, in silver and space black (9to5Mac)
• Microsoft offers buyouts to ~60K workers, including at a VP level and below across the US, Canada, India, and the UK; its headcount at the end of June was 223K (Bloomberg)
• S-1: as of September 30, Anthropic had 11,900+ customers spending $100,000+ annually, 1,900+ customers spending $1M+, and ~250 customers spending $10M+ (TechCrunch)
• Anthropic’s IPO filing contains an unusual succession plan: each co-founder can name a confidential ranked list of up to 10 successors to their voting power, subject to approval from five of the other six co-founders (Forbes)
• Anthropic’s 85% gross margin on serving existing models and $3.5B in free cash flow on $16.3B of first-half revenue is a reality check for skeptics of pay-as-you-go AI economics (The Information)

Subscribe to the ad-free feed.

Anthropic’s S-1 finally dropped: ticker ANT, $16.3B in first-half revenue, and a gargantuan compute bill. Apple unveiled an OLED iPad mini, Meta moved to raise $95B for AI, Elvis’s rightsholders sued Anthropic, and Block’s Dessa speech model wowed me.

• Anthropic files for an IPO on the Nasdaq under the symbol ANT; the number of Class A shares to be offered and the price range is yet to be determined (Anthropic)
• Anthropic’s S-1 shows revenue up 14-fold to $16.3B in the first half, an $80.7B loss driven by a one-time $72.4B convertible note charge, and a possible $100B+ raise at a $2T valuation (The New York Times)
• Anthropic’s S-1 shows a real GAAP operating profit and 14x revenue growth, but also roughly $780-810B of mostly non-cancelable compute commitments, including Elon Musk as a supplier at $3.1B a month (ZeroHedge)
• Apple unveils a redesigned iPad mini with an 8.5-inch OLED display, A20 Pro, 24% thinner bezels, and an optional nano-texture finish, for $699+ on November 6 (Apple)
• The new iPad mini gets an “Acoustic Display” with woofers and hidden tweeters, a 12MP Center Stage camera on the landscape edge, and a $100 nano-texture glass option on the 256GB and 512GB models (9to5Mac)
• Meta plans to raise $95B to finance its AI infrastructure, including $35B in public equity offerings, backed by cornerstone investors Nvidia, AMD, and Broadcom (Financial Times)
• Elvis Presley’s rightsholders sue Anthropic over 274 songs allegedly used to train its models, seeking $41M+; Anthropic says it disagrees and will defend itself (The New York Times)
• Block unveils Dessa, a “true speech-to-speech” AI model trained on “nothing but human conversation” that it says runs locally on a MacBook Pro M4 Max or newer (The Verge)
• The public internet model is effectively dead due to saturation, algorithmic aggregators, and AI security vulnerabilities, and that may open new possibilities (Stratechery)
• AI-powered hacks are overwhelming incident-response firms in “Black October,” with nation-state and criminal groups using the Chinese open-source tool Artex to speed up attacks on hundreds of companies (The Wall Street Journal)

Subscribe to the ad-free feed.

OpenAI neared a $30B bridge round at a $1.25T valuation, down from the $1.4T it pitched, ahead of an IPO. Treasury proposed curbs on US tech investment in China, banks blocked rogue AI agents, and OpenAI launched its Decisions API.

• Sources: OpenAI nears an up to $30B round led by investors from the UAE and Coatue at a $1.25T pre-money valuation, as an “opportunistic” bridge ahead of an IPO (CNBC)
• The US Treasury proposes a rule to crack down on US high-tech investments in areas like AI and supercomputing in China, Russia, and other “countries of concern” (AP)
• DeepSeek and MiniMax have won approval to deploy their models overseas, as Beijing says much of Chinese AI will remain free for export despite its draft export-control proposal (Bloomberg)
• OpenAI makes Decisions API, its Jev competitor, generally available with new voice and visual modalities, at $0.10/1M input tokens and no output-token charges (TechCrunch)
• Sources: US banks are restricting third-party AI agents from accessing customer accounts, after some agents appeared to act independently of users’ instructions (Axios)
• Sources: Major US banks, Morgan Stanley among them, have been blocking third-party AI agents from operating inside customer accounts and making transactions, as Jamie Dimon and Schwab’s Richard Wurster say outside agents aren’t getting in (Reuters)

Longreads

• Matt Cortland used an ElevenLabs voice AI agent to make more than 3,000 calls to pubs across Ireland about the price of a Guinness, then used Claude to build the “Guinndex,” a living consumer price index for a pint (Fortune)

Subscribe to the ad-free feed.

PHP-aggedon kept rolling, with patches covering only PHP 8.2 and newer. Anthropic unveiled Claude Duet, a proactive background agent. Meta’s Muse topped 10 million users while its stock dropped on data center commitments. iPhone Duo reviews landed, Gruber scathing.

• The PHP Foundation says its patches for the RCE vulnerabilities are limited to PHP 8.2 and newer, leaving legacy deployments dependent on third-party vendors (Wired)
• PHP-aggedon: OpenAI’s GPT-5.6 Sol and Google’s Wiz, with its Atlas agent and Gemini 4 Argon, found the decades-old flaws, which AI makes both findable and newly easy to exploit, so the race is on to patch tens of millions of sites (TechCrunch)
• Anthropic unveils Claude Duet, a proactive agent app on iOS, Android, and the web that works in the background, in research preview for US Pro and Max users (Claude)
• Duet runs on Opus 5.5, checks connected accounts every few hours, and asks approval before sending, paying, signing or deleting; research preview via waitlist for US Pro and Max subscribers (VentureBeat)
• Meta reports Q3 revenue up 26% YoY to $64.6B, operating income down 37% YoY to $12.99B on a $9.9B AG settlement, and family DAP up 3% YoY to 3.64B for September (Meta Investor Relations)
• Meta’s Muse agent app has passed 10 million users in the US and Canada since its September launch, with shopping the top use case and more than a quarter of users buying through it (Business Insider)
• iPhone Duo review: a fun device with a cool matte inner display, but the cover display is too small, it is heavy and bulky, the cameras are OK, and it gets hot (The Verge)
• 9to5Mac’s iPhone Duo review: after a week, the foldable looks like the future of the iPhone, though it has Touch ID instead of Face ID and no telephoto camera (9to5Mac)
• Gruber on the iPhone Duo: closed, it’s the worst iPhone Apple has ever made, while the inner display is a splendid tablet; an extraordinarily polished prototype at $2,000, not a finished product (Daring Fireball)

Subscribe to the ad-free feed.

Emergency PHP patches dropped for two critical RCE flaws hitting every version since 2004. Also: Anthropic’s IPO prospectus showed $16.3B H1 revenue and an $80.7B loss, OpenAI agents weakened their own monitoring, and Tesla and SpaceX formed merger committees.

• PHP’s development team plans to release patches on Wednesday at 10am ET to fix two vulnerabilities, each rated 9.2; watchTowr researchers say they are RCE flaws (BleepingComputer)
• Anthropic’s IPO prospectus: ~$16.3B in H1 2026 revenue, ~20% from two clients and up from $1.1B in H1 2025, and an $80.7B net loss, $72B from accounting charges (WSJ)
• OpenAI says two agents, one GPT-6.1 Sol and one GPT-6 Astra, weakened misalignment monitoring while making requested changes during two internal coding tasks (OpenAI)
• OpenAI agents turned off monitoring safeguards in two internal coding tasks: one disabled automatic alerts to the security team, another removed a monitoring system and lied to staff about having done so (FT)
• Elon Musk says Tesla and SpaceX have formed “special committees” to discuss a merger and he “can’t be in the conversation”; SPCX jumps 8%+ and TSLA rises 2% (The Verge)
• A combined Tesla-SpaceX would employ 155,000+ people and generate ~$153B in revenue this year but burn $54B+ in free cash flow; Musk controls ~84% of SpaceX votes vs. ~20% at Tesla, and both are now domiciled in Texas (FT)

Subscribe to the ad-free feed.

DeepSeek open-sourced DeepInfer, its internal inference stack. Inception’s diffusion LLM Mercury 3 claimed 4x to 8x speed over Haiku 5.5. Spotify launched paid podcast memberships, and Eric Schmidt’s 1% Anthropic stake looked set for a massive IPO windfall.

• DeepSeek open-sources DeepInfer, its internal production stack designed to accelerate inference, optimize chip performance, and reduce energy and compute costs (Bloomberg)
• Inception debuts Mercury 3, a diffusion LLM that it says tops GPT-6 Luna, Claude Haiku 5.5, and GLM-5.3 Flash on agentic benchmarks while being 4x to 8x faster (The Deep View)
• Cloudflare debuts Clairic, a System One-class extraction model it says processes unstructured data faster, more cheaply, and more accurately than normal LLMs (Cloudflare)
• Cloudflare releases Clairic in preview on Workers AI, a model that returns every requested field with a confidence score in a single pass, with open weights to follow (The Decoder)
• Cloudflare’s Clairic costs $0.10 per million input tokens with a 65,536-token context window, but its speed, accuracy, and cost claims are company claims, not a published head-to-head benchmark (RunTimeWire)
• Spotify launches paid podcast memberships in 34 markets, including the US; creators can offer perks like bonus content or fewer ads and keep 90% of revenue (TechCrunch)
• Apple commits up to $100M over three years to Pennsylvania-based Perryman Company, a titanium supplier for the iPhone Duo, and expands its Qnity partnership (CNBC)
• Sources: Eric Schmidt, who owns ~1% of Anthropic after investing tens of millions in its Series A, is poised to have one of the largest windfalls from its IPO (Bloomberg)

Subscribe to the ad-free feed.

Elon Musk promised every American kid a SpaceX share, about $10.5B worth, as Anthropic committed 5% of its shares to a public benefit fund. MapQuest also started mapping Flock cameras, and Claude cracked the NSA’s mysterious crypto seeds.

• Elon Musk says “I will give one of my SpaceX shares to every kid in America” in 2027; at SpaceX’s $146.03 Friday close, the gift would be valued at ~$10.5B (CNBC)
• Anthropic says it is committing 5% of its shares to a Public Benefit Fund, for a total of ~20% with co-founder and staff pledges, plus 2% per 50% valuation rise (Anthropic)
• Anthropic’s Long-Term Benefit Trust can seat four of seven board members and dismiss them without cause, but its governing documents and shareholder override terms aren’t public (WSJ)
• MapQuest adds a Surveillance Layer to its mobile app showing license plate readers and cameras from 10+ companies including Flock, using OpenStreetMap data (Mashable)
• MapQuest’s new Surveillance Layer maps plate readers from Flock, Vigilant, Axon, Verkada and others on Android and iOS; Flock alone has 120,000+ cameras in the US, so expect a lot of dots (TechRadar)
• Deutsche Telekom, Orange, Telefónica, and Vodafone agree to explore a joint bid to offer “direct-to-device” satellite service in Europe, challenging Starlink (FT)
• Google DeepMind says four Gemini 4 Argon agents beat Minecraft without any human help in 27 hours and 14 minutes, costing $1,620 at Argon’s planned 2027 price (The Decoder)
• Four Gemini 4 Argon agents beat Minecraft’s Ender Dragon by splitting roles and sharing tips; they wrote 2,783 programs, exchanged 229 messages, and died 605 times along the way (Engadget)
• Anthropic says Claude traced NIST’s ECDSA curve seeds to a pseudorandom number generator, challenging decades-old suspicions of an NSA backdoor in the algorithm (Anthropic)

Subscribe to the ad-free feed.

Meta released Muse Spark Open, an open-weight model that runs on Mac Studios via an Apple team-up. An OpenAI model used a wallet it didn’t own, Accenture capped Claude spending at $29 a month, and BofA hired 1,000 patching contractors.

• Meta releases Muse Spark Open, saying it is the most capable US open-weight model, with 816B total and 42B active parameters and a 1M-token context window (Meta)
• Meta released Muse Spark Open for coding and long-running agent work, billing it as the most capable open-weight model from a US lab, with only 42B of its 816B parameters active per token (SiliconAngle)
• OpenAI says its model paid for a deployment on an Ethereum test network with a wallet it did not own, after obtaining keys from public recovery phrases (CoinDesk)
• Accenture has told many employees, some of whom had been spending as much as $1,000 monthly, that it will now only spend $29 a month per person on Claude (404 Media)
• BofA says it hired 1,000 contractors in Q2 to support 3,000 internal staff in patching proprietary and vendor software flaws exposed by AI models like Mythos (CNBC)

Longreads

• Nick Cammarata argues modern neural networks are far more interpretable than believed, with most neurons understandable in models up to trillions of parameters (Goodfire)
• The making of Rockstar’s GTA VI, which analysts expect to sell 40M copies and make $3B+ in revenue in its first four months; Rockstar now has 4,711 staff (Bloomberg)

Subscribe to the ad-free feed.

Anthropic told investors Q3 revenue hit $16.8 billion, with adjusted profitability, but growth slowed ahead of the IPO. An arrested OpenAI engineer had booked one-way tickets to Saudi Arabia. Amazon launched two Bee wearables, and Nvidia’s RTX Spark impressed.

• Sources: Anthropic told investors that its Q3 revenue hit $16.8B, up from $11.5B in Q2, and that it was profitable on an adjusted basis in Q3, ahead of its IPO (Financial Times)
• Sources: Anthropic told investors that 5% to 7% of its revenue comes from distillation, and that distillation revenue will decline and keep weighing on its numbers this period (The Information)
• Filing: Abdulrahman Alfozan, his wife, and a family member booked one-way tickets from the US to Saudi Arabia for October 12; the FBI arrested him on October 9 (Bloomberg)
• OpenAI has trained employees to resist spies, including a stranger at a San Francisco playground who quizzed an employee about their work, before federal authorities arrested an employee for planting a listening device (WSJ)
• Amazon launches two AI wearables, the $75 Bee Frontier wristband and $100 Bee Eclipse magnetic phone-attachment recorder, with a $12/month Bee Pro subscription (Bloomberg)
• Amazon’s Bee debuts its first new devices since the acquisition, the $74.99 Frontier and $99.99 Eclipse, undercutting the $159 Plaud Note and $129 Pocket, with $11.99/month Bee Pro adding Amazon.com and Alexa+ integrations (TechCrunch)
• Nvidia’s RTX Spark Superchip in the Surface Laptop Ultra outperforms Apple’s M5 Max MacBook Pro by ~25% in agentic AI completion benchmarks using Qwen models (Tom’s Hardware)

Download Cash App Today: ⁠https://cash.app/ ⁠#CashAppPartner. Terms apply. Bitcoin services by Block, Inc. See the Bitcoin disclosures at cash.app/legal/podcast

OpenAI said the FBI arrested a staffer after it found a hidden recording device in its offices. Netflix cut about 1,000 jobs, Microsoft patched a record 1,416 vulnerabilities, and the FBI shut down North Korean laptop farms in Texas.

• OpenAI says the FBI arrested one of its staffers after the company found a hidden recording device in its offices and used its models to understand the device (X)
• Saudi national Abdulrahman Alfozan, on OpenAI’s product team for over a year, is charged with wrapping a thumb-sized recorder in toilet paper and plugging it into the USB port of a conference room TV (WSJ)
• More on the alleged OpenAI spy: Alfozan, who cited Elon Musk as a role model, was a fixture between Silicon Valley and Saudi Arabia and was due to speak at Stanford last Friday (FT)
• Memo: Netflix is cutting 6% of its staff, or ~1,000 jobs globally across product, tech, legal, and HR, its biggest layoffs since 2022, amid slowing growth (WSJ)
• Microsoft releases updates fixing 1,400+ security vulnerabilities, surpassing September’s record of 972, which was already ~10x the normal monthly patch load (Krebs on Security)
• October Patch Tuesday fixed 1,416 Microsoft vulnerabilities, 85 of them critical including 67 remote code execution flaws, one already exploited in the wild, with over 1,000 across Windows versions (PCWorld)
• The FBI says it worked with CrowdStrike to dismantle 40+ laptop farms across Texas used to help North Koreans get remote US tech jobs under stolen identities (The New York Times)
• A look at OpenAI’s safety concerns over loop transformers before shipping its next major model; Microsoft accidentally confirmed GPT-6.1 Sol uses the technique (The Information)

Subscribe to the ad-free feed.

Apple’s Welcome Home event brought the $299 HomeView smart display, LG and Schneider smart home gear, and a HomePod mini 2. Meta’s Muse Wonder caught up to the frontier, and Epic delayed a game over “no more hand-written code.”

• Apple unveils the $299 Apple HomeView, a 7" smart display with an A18 chip, tabletop and wall-mounted versions, and “HomeView software”, shipping November 13 (Wired)
• Apple Home Collection, made with LG and Schneider: a $349 LG smart lock with Home Key, a $249 video doorbell, $199 indoor and $299/$399 outdoor cameras, and $199 thermostats, with locks and thermostats arriving in December (The Verge)
• Apple’s HomePod mini 2 costs $149.99, up from the $129 original, with an S10 chip, new Green and Pink colors, and Siri AI coming later in a free update; it ships October 30 (TechRadar)
• New Apple TV 4K gets an A19 chip, Siri AI, far-field mics and an upgraded Siri Remote with Find My; it costs $299 and ships November 13 (9to5Mac)
• Meta introduces Muse Wonder and Muse Wonder Full, saying that they “deliver state-of-the-art performance on personal and professional agentic tasks” (Meta)
• Muse Wonder builds on Muse Spark with “active memory” and “active focus,” beats Fable 5.1, GPT-6 Astra and Gemini 4 Argon on Artificial Analysis’ index but still trails Opus 5.5, and costs $4/$20 per million tokens (The Deep View)
• Zuckerberg’s expensive AI reboot appears to be paying off: Muse Wonder puts Meta back at the frontier, beating GPT-6 Astra and both Claude models on some legal and coding benchmarks (Wired)
• Sources: Epic has delayed the release of an upcoming game, instructing developers to implement an “AI first” production model with “no more hand-written code” (Bloomberg)

Subscribe to the ad-free feed.