The Automated Daily - AI News Edition

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Lindy is your ultimate AI assistant that proactively manages your inbox - https://try.lindy.ai/tad - KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad - Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad Today's topics: Open models cutting AI coding costs - AI coding tools are increasingly judged on token spend and ROI, and vendors claim frontier open-weight models can handle a large share of routine software engineering tasks at lower cost. Model routing between open and closed - Intelligent model routing sends simpler coding requests to cheaper open models and harder ones to closed frontier LLMs. Per-user and per-API-key budget tracking is becoming standard in enterprise AI governance. Treat vendor benchmarks with caution - Vendor-published benchmarks comparing open and closed LLMs on cost and speed should be independently verified. Engineering teams are advised to test model routing on their own codebases and workloads. - Fireworks Nexus Promises Cheaper Frontier AI for Coding Teams Episode Transcript Open models cutting AI coding costs Here's the idea making the rounds. AI coding assistants have become routine for engineering teams, and the bills have grown with them. Every autocomplete, refactor, and agent run burns tokens, and when those tokens come from top-tier closed models, the costs add up quickly. So a growing number of platforms are pitching a fix: keep developers working exactly as they do now, but quietly send many of those requests to open-weight models instead. The latest example is a code-generation offering from inference provider Fireworks, which argues that frontier open models can now handle a large share of everyday engineering work. Model routing between open and closed The core mechanism is called model routing. Rather than sending every request to one premium model, a routing layer decides on the fly which model fits the job. Simple tasks, like boilerplate or small edits, go to a cheaper open model. Harder problems, like complex debugging or architectural changes, can still go to a closed frontier model. Why it matters: if routing works well, teams pay premium prices only when they truly need premium performance. Alongside that, these tools increasingly offer finance-friendly features, such as tracking spending by individual developer, by model, or by API key, and setting hard limits. That signals AI coding is maturing from an experiment into a line item that procurement and finance teams want to manage closely. Treat vendor benchmarks with caution A word of caution, though. The cost and speed comparisons in pitches like this usually come from the vendors themselves, and benchmark results don't always carry over to a real company's messy codebase. Open models have improved dramatically, but the gap with the best closed models still shows up on harder, multi-step tasks. The sensible approach for engineering leaders is to run their own trials: route a slice of real work to open models, measure quality and developer satisfaction, and only then decide how far to shift. Still, the broader direction is clear. Competition from open models is putting real pressure on the price of AI coding, and that's good news for anyone paying the bill. Subscribe to edition specific feeds: - Space news * Apple Podcast English * Spotify English * RSS English Spanish French - Top news * Apple Podcast English Spanish French * Spotify English Spanish French * RSS English Spanish French - Tech news * Apple Podcast English Spanish French * Spotify English Spanish Spanish * RSS English Spanish French - Hacker news * Apple Podcast English Spanish French * Spotify English Spanish French * RSS English Spanish French - AI news * Apple Podcast English Spanish...

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad - Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron Today's topics: AI pricing models under pressure - AI and SaaS companies are moving away from traditional subscription pricing toward usage-based and outcome-based monetization. Leaders from Replit, Snowflake, Clay and Fin discussed the shift at a Stripe and Metronome event. Usage-based billing and spend controls - Usage-based AI products need customer spend controls and budget caps to avoid surprise bills. Clear limits help build trust and support pricing experiments in AI software. Protecting margins as compute costs swing - Inference and compute costs make AI gross margins hard to predict. Cost forecasting and pricing tied to customer value are becoming core strategy for AI startups. Sales compensation in usage-based AI - When revenue depends on consumption instead of signed contracts, sales compensation, quotas and finance processes must change. Product, finance and sales teams have to adapt pricing together. - Stripe Monetize 2026 On-Demand Talks on AI Monetization Episode Transcript AI pricing models under pressure The spark for today's discussion is a set of talks from a monetization conference run by Stripe and Metronome. We'll skip the event itself and focus on what was said. The speakers included people from Replit, Fin, Snowflake, and Clay, plus the author of the Growth Unhinged newsletter. They kept coming back to the same point: AI products don't fit the old software pricing playbook. For about twenty years, SaaS was simple. You paid per seat, per month, and the cost of serving one more user was close to zero. AI changes that. Each request uses real compute, so a heavy user can cost far more than a light one. That's why companies are moving to usage-based pricing, credits, and in some cases pricing tied to results. Usage-based billing and spend controls Usage-based pricing creates its own trust problem. If customers can't predict their bill, they hesitate to use the product, or they get a surprise invoice and leave. Spend controls were a major topic for that reason. Budget caps, alerts, and clear limits are no longer optional extras. They're part of what customers expect before they let an AI agent run inside their business. The speakers also talked about pricing experiments. Many teams admit they're still searching for the right model, and they're testing different structures in the market instead of committing to one approach for years. Protecting margins as compute costs swing Margins are the other side of the issue. The costs of running AI models move around. Model prices drop, usage spikes without warning, and new features can quietly double compute costs. So cost forecasting is now a core skill, not just a spreadsheet task for finance. The underlying question is how to measure the value a customer actually gets, and then charge in a way that matches it, without giving away margin on the heaviest users. This is a big reason some AI companies report strong revenue growth while still facing hard questions about long-term profitability. Sales compensation in usage-based AI The last theme got less attention but it matters inside these companies: sales compensation. When revenue depends on how much customers use the product, not on the size of a contract they signed, the old way of paying salespeople stops working. A large deal can look good on paper and then bring in little if adoption stalls. Companies are rethinking quotas and incentives to reward actual consumption and customer success. The broader takeaway is that a pricing change...

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Consensus: AI for Research. Get a free month - https://get.consensus.app/automated_daily - Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad Today's topics: DeepMind AI verifies Signal protocol - Google DeepMind's AlphaProof Code used AI agents and the Lean proof assistant to formally verify security properties of the real Signal protocol implementation. The project generated over six million lines of Lean in four weeks, and DeepMind is open-sourcing the verification artifacts. Google launches Gemini 4 Carbon - Google released Gemini 4 Carbon, its most capable frontier model so far, aimed at agentic workflows, software engineering, cybersecurity and scientific reasoning. It is available through AI Studio, Antigravity's multi-agent /swarm feature, the Gemini app and AI Mode in Search. Microsoft MAI-Transcribe-2 speech model - Microsoft AI introduced MAI-Transcribe-2, a speech recognition model covering 60 languages with speaker diarization and word-level timestamps. Microsoft says it leads the FLEURS benchmark and costs ten cents per hour of audio. Forecasters shorten AI timelines to 2027 - AI forecasters cite new METR evidence that AI is speeding up AI research by about 1.7x, and now expect a Superhuman AI Researcher around 2027. They argue this shorter timeline justifies pausing or substantially slowing frontier AI development. - Buildkite Updates Pricing with All-Access Trial, Tiered Plans, and Usage-Based Billing - AI Futures Project: Timelines Shorten, SAR Now Expected in 2027 - Microsoft Launches MAI-Transcribe-2 Speech Recognition Model - DeepMind Says AlphaProof Code Formally Verified Signal Protocol - Buildkite Promotes AI-Ready CI Platform for Large-Scale Workflows - Google releases Gemini 4 Carbon Episode Transcript DeepMind AI verifies Signal protocol We'll start with Google DeepMind, which says it used a new agentic system called AlphaProof Code to formally verify key security properties of the Signal protocol. The important detail is that the proofs cover the real, deployed implementation, not just the design on paper. Plenty of cryptographic designs have been checked mathematically, but their implementations are where bugs usually appear, and DeepMind says no one had formally verified an implementation like this by any method until now. Human experts wrote the specifications of what 'secure' means. AI agents then produced the proofs in Lean, a proof assistant that checks every step on its own. In four weeks the system produced more than six million lines of Lean, and humans needed to review only about one percent of it. As a test, the team put two old, already-fixed Signal bugs back into an isolated copy of the code, and the system correctly showed that this version was insecure. DeepMind is open-sourcing the artifacts and sees the same approach applying to medical devices and power grids. Until now, formal verification has been too slow and expensive for most real-world code. If AI handles the routine proof work, that could start to change. Google launches Gemini 4 Carbon Google also had a second announcement this week. Gemini 4 Carbon, released on November 10th, is the company's newest frontier model, and Google calls it its strongest so far. It is aimed at hard knowledge work: long-running agentic tasks, software engineering, cybersecurity and scientific reasoning. Google reports top results on several coding and automation benchmarks and says it costs less per task than competing models. Carbon is available almost everywhere at once, including AI Studio for developers, Gemini for business, the Gemini app and AI Mode in Search. Scientists and...

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad - Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad - Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily Today's topics: METR: AI Doubling Anthropic's R&D Speed - A new METR report estimates AI-assisted AI research at Anthropic is speeding up model progress about twofold, raising questions about recursive self-improvement and AI safety timelines. Anthropic's Compute Bill and Gas Turbines - Anthropic's amended S-1 shows roughly $895 billion in compute commitments, and the company is reportedly spending $3.4 billion on Caterpillar natural gas turbines to power its data centers. Claude Fable and Mythos 5.2, Opus Complaints - Anthropic released Claude Fable and Mythos 5.2 with coding, reasoning and vision gains, while some users claim Opus 5.5 in high mode now reasons less deeply in Claude Code. Meta Muse Ember Efficient Agent Model - Meta launched Muse Ember, a small, low-cost model with a 1M-token context window built for agentic tool calls, coding subtasks and document parsing. Google Unveils Two New TPUs - Google announced TPU 8tx for confidential inference and TPU 8i Max for long-context agentic workloads, expanding its AI hardware lineup with PyTorch support. Smarter Compression and Test-Time Compute - AutoCompress tailors LLM compression to real GPU latency and memory limits, while new essays examine compute extrapolation and reasoning as search through token space. US Could Import AI Tokens - An analysis warns US AI token demand could outpace domestic data center and power supply between 2028 and 2030, making the US a net importer of AI compute. NVIDIA GR00T 2 Robot Foundation Model - NVIDIA's GR00T 2 predicts video and robot actions together, learning from human egocentric video and transferring skills to robots across different bodies. Apple Harbor and Paid Logic Pro AI - Apple is reportedly building Harbor, a local AI inference server app that pools Mac memory over Thunderbolt 5, and now sells paid AI usage packs in Logic Pro. OpenAI Control Room Kill Switch - OpenAI built Control Room, a centralized system that can quickly halt model workloads across research infrastructure for AI safety and security, with fail-closed design. Fireworks AI Breach Grows Wider - Fireworks AI says its October security incident also exposed account, billing and some customer data, linking it to an LLMJacking cybercriminal group. Open Source Projects Rejecting AI Code - The Open Source Says No catalog tracks projects like QEMU, Zig and Servo that ban or restrict LLM-generated contributions, documenting the open source backlash. AI Reshaping News, Freelancing, Workdays - Studies show generative AI becoming a news gateway amid audience skepticism, AI reshaping Upwork freelance projects, and AI users working longer, more fragmented days. - NVIDIA Unveils GR00T 2 Robot Foundation Model - Granola Launches AI Notepad for Meeting Notes and Follow-Ups - Open Source Projects That Ban AI Contributions - Anthropic’s $895 Billion Compute Bet Puts Growth to the Test - Claim That Opus 5.5 Was Quietly Nered - Meta Launches Muse Ember, a Low-Cost AI Model for Agentic Work - ChatGPT Enterprise Launches in AWS Marketplace - OpenAI Launches Control Room to Stop Research Model Workloads Quickly - AI Is Becoming a News Gateway, but Audiences Still Reject AI-Made Journalism - Apple starts selling extra AI access for Logic Pro - LiveKit Announces Introductory Voice Agent Workshop - AutoCompress: Hardware-Guided LLM Compression - Fireworks AI Expands Disclosure of October 2026 Security Incident - Wiz Explains How to Secure AI Agents - Apple Reportedly Developing Harbor Mac App for Local AI Models - METR Report...

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Lindy is your ultimate AI assistant that proactively manages your inbox - https://try.lindy.ai/tad - Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad Today's topics: AI-Proven Quantum Computing Conjecture - PSI says its AI research system, running on a non-public Anthropic model and checked with the Lean proof assistant, proved the Aaronson-Ambainis conjecture. The conjecture is a long-standing open problem in quantum complexity theory about when quantum computers can be exponentially faster than classical ones. AI Accelerating AI Research - Ajeya Cotra reads a new METR report as evidence that AI is already doubling the pace of AI research at Anthropic. A separate analysis suggests OpenAI's internal AI R&D time horizons may be doubling roughly every two months. AI Agents Redesigning Networks Mid-Training - Researchers placed an AI coding agent inside a neural network's training loop so it could change the model's architecture while it learns. The approach, called mechanistic intervention, reduced validation loss and improved results on reasoning tasks. OpenAI BioBasicsBench Lab Results - OpenAI's BioBasicsBench tests AI agents on routine wet-lab tasks using raw experimental data. GPT-6 Astra scored 66.7%, ahead of GPT-5.6 Sol at 43.8%, but it still made errors of scientific judgment. Warp Sciences AI Drug Discovery - Kevin Weil launched Warp Sciences with a $215 million Series A led by Lightspeed. The company plans to build an AI- and robotics-powered preclinical CRO in the US to speed up medicine discovery. Anthropic Project EVE Game Agents - Anthropic's Project EVE tested Claude agents in EVE Online and EVE Frontier, covering market trading and team coordination. The agents earned steady trading profits but rarely noticed when a teammate was secretly sabotaging them. Anthropic Layered Agent Memory Research - Anthropic describes agent memory as a layered system of transcripts, notes and consolidated knowledge. Experiments on a database benchmark and in Pokémon Red showed that selective memory and offline 'dreaming' improve long-running AI agents. GitHub Next Ace Multiplayer Agents - GitHub Next released Ace, an experimental open-source tool for collaborative, local-first agent workflows. Teammates share channels while agents run directly on the host's Mac with that user's privileges. AI Coding and Technical Debt - An essay and a Reddit thread argue that heavy reliance on AI coding tools inflates codebases and increases technical debt. They also point to weaker code review and lost engineering skills when human oversight fades. Notion AI Fireworks Security Incident - Notion says a breach at its vendor Fireworks AI may have exposed unattributed Notion AI chat transcripts from fewer than 1% of workspaces. Notion cut ties with Fireworks and recommends rotating any credentials shared in those chats. OpenAI Funding and CoreWeave Bonds - OpenAI reportedly raised $11 billion at a $1.25 trillion valuation while its IPO plans are delayed. Meanwhile, a CoreWeave 2032 bond yield rose above 16%, a sign of growing credit stress in AI infrastructure financing. Meta Muse Adds Custom Spaces - Meta added 'spaces' to its Muse app, letting users describe and build custom interactive experiences such as planners, trackers and games. The update moves Muse toward a personalized, user-built AI workspace. Rethinking Firms and AI Futures - John Nay argues for building AI-native firms like Norm Law instead of adding AI to old workflows. Patrick Collison asks what a positive mature AI society would look like and raises questions about AI moral worth. - Buildkite Updates Pricing with All-Access...

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad - Consensus: AI for Research. Get a free month - https://get.consensus.app/automated_daily Today's topics: Is the AI bubble real? - An analysis argues the AI industry may not be in a financial bubble, citing Anthropic revenue growth, Nvidia, Meta and Alphabet earnings, and data center spending as one-time capital costs. Apple trims engineers' AI budgets - Apple reportedly cut some engineers' daily AI spending cap from $300 to $200, highlighting enterprise generative AI costs and the growing focus on efficiency, model routing and caching. Managing coding agents with git queues - A developer organizes Claude Code and Codex work as a git-based priority queue of markdown tickets, reducing context switching and cognitive overload in AI-assisted coding. Engineers acting as 'meat proxies' - A software engineer argues that forwarding tasks to AI agents without review, the so-called 'meat proxy' behavior, is serious misconduct that shifts risk and cleanup onto teammates. Asahi Linux relicenses over AI laundering - Asahi Linux is switching from the MIT license to GPL v2-or-later after unreleased code was allegedly copied and laundered with LLMs, a notable open-source response to AI code theft. Etsy's one-sided AI training terms - Etsy's updated terms appear to let it train AI on creators' work while banning users from doing the same with Etsy content, fueling backlash over intellectual property and AI. Author publishes 370 AI-assisted books - Romance writer Coral Hart says she published over 370 books in 20 months using AI models, sparking debate over AI slop, authorship, creativity and the future of publishing. Bird-mimicking drone evades Air Force sensors - Ornadyne's O1 ornithopter drone flaps its wings and mimics local bird species, registering as a bird on sensors during an Air Force exercise at Edwards Air Force Base. - Why the AI Bubble May Not Exist - How the Author Uses Git and a Priority Queue to Manage AI Work - Apple Cuts Engineers’ Daily AI Budget to $200 ([stork.ai](https://www.stork.ai/newsletter/2026-11-08-apple-capped-engineers-at-a-200-daily-ai-budget)) - Bird-Like Drone Evades Detection in Air Force Test - Asahi Linux Switches to GPL to Fight AI-Driven Code Laundering - Why “Meat Proxying” AI Work Should Be Punished - South African Author Defends Publishing Hundreds of AI-Assisted Books - Etsy Faces Backlash Over AI Training Terms Episode Transcript Is the AI bubble real? Let's start with money. A widely shared essay argues that the AI industry may not be in a bubble, or at least not one that's about to pop. The author's case is that the huge sums going into data centers are mostly one-time capital investments, not permanent operating costs that can never be recovered. He points to fast revenue growth at Anthropic and to solid earnings and fairly ordinary valuation ratios at Alphabet, Meta and Nvidia as signs of real business value. Even without strong moats, he says, brand loyalty and institutional lock-in can keep the money flowing. He does acknowledge real risks, such as regulation, public backlash and slower growth, but concludes that much of the bubble talk is closer to wishful thinking than financial analysis. Apple trims engineers' AI budgets That said, costs are clearly on people's minds, even at the richest companies. Apple has reportedly lowered the daily AI spending cap for some engineers from three hundred dollars to two hundred. It's a small number for a company of Apple's size, but it shows how quickly model usage can add up when it's left uncapped across large teams. The takeaway from industry watchers is...

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Invest Like the Pros with StockMVP - https://www.stock-mvp.com/?via=ron - KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad - Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily Today's topics: Microsoft shuts AI Frontiers research - John Langford says Microsoft's AI Frontiers group has ended after being moved into Microsoft AI, citing a leadership focus on scaling over fundamental research. Internships will not continue, raising questions about Microsoft Research's role. Anthropic warns of recursive self-improvement - In a revised S-1 filing, Anthropic says it may soon reach recursive self-improvement, where AI accelerates AI research faster than safety and alignment processes can keep up. It says no current Claude model has crossed that threshold. AI and software engineering jobs - An analysis argues the software hiring slump was driven by interest rates and the pandemic hiring boom, not ChatGPT. AI's clearest effect is weaker hiring for junior developers, while senior and AI-related roles grow. Judging AI versus human content - An essay argues AI-generated web content isn't fundamentally worse than human content, since both can be inaccurate or bland. It supports AI disclosure as a courtesy but urges skepticism toward all online information. 0rwx labels for AI involvement - The 0rwx framework borrows Unix-style permissions to describe AI's role in creative and technical work, from no AI access to fully AI-generated output. It aims to clarify authorship, control, and vibe coding discussions. Brits stay polite with AI - A survey reported by TechRadar finds 70% of Brits say please or thank you to voice AI assistants, and some even apologize. The habit shows how AI is becoming normalized as a social partner. Silicon metasurface focuses ultraviolet light - University of Stuttgart researchers built a flat silicon metasurface using air-filled cavities to focus ultraviolet light near the diffraction limit. Its single-step fabrication could aid compact UV imaging, spectroscopy, and sensing. Stunt Car Racer opponent AI explained - A deep dive explains how computer-controlled opponents in the classic Commodore 64 game Stunt Car Racer drive, swerve, block, and collide. It reveals hidden mechanics behind retro game AI behavior. - John Langford Says Microsoft’s AI Frontiers Is Over - Why the Author Doesn’t Care if Web Content Is AI-Generated - Air Holes in Silicon Create Flat UV-Focusing Mirrors - 0rwx Proposes a Four-Level Model for AI Use - Brits Are Surprisingly Polite to AI, Survey Finds - Anthropic Says It May Be Close to Recursive Self-Improvement - AI Fear vs. the Software Job Market - How Stunt Car Racer Opponents Behave Episode Transcript Microsoft shuts AI Frontiers research Let's start with that Anthropic update. In the story we've been following around its IPO paperwork, Anthropic's revised S-1 filing now says the company may soon reach what it calls recursive self-improvement, the point where AI meaningfully speeds up the creation of more capable AI. The company warns that leaning on AI in its own research could push progress faster than its safety, alignment, and security processes can handle. Its internal projections suggest it may imminently hit its own threshold for dramatically accelerated, AI-driven progress, though it says no current model has crossed that line yet. Anthropic also stressed it doesn't expect to let Claude design successive frontier models on its own without real human oversight. Seeing this kind of language in a securities filing, not a blog post, is what makes it notable. Anthropic warns of recursive self-improvement Staying with the big labs, there's a sobering note from inside Microsoft. Researcher John Langford writes that...

Support The Automated Daily: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support This Week's Topics: Voters, markets and AI debt - In the US midterms, voters passed dozens of local measures restricting data centers, the huge computer facilities behind AI, over electricity bills, water use and environmental strain. The next day the Nasdaq fell roughly five and a half percent, with fears that a Democratic Congress could slow the data center boom. Chip stocks ended down nearly thirty-six percent, though the broader market looked bruised rather than broken. Before the vote, Apollo's chief economist found little AI payoff outside tech, big tech borrowing kept growing, off-balance-sheet AI commitments reportedly neared 3.2 trillion dollars, and Oracle's shares fell after it moved to raise twenty billion dollars. AI demand is even causing shortages of circuit boards used in ordinary computers. AI speeds up its own making - Anthropic's stock-market filing warns it may soon reach recursive self-improvement, where AI meaningfully speeds up building better AI, and says progress could outpace its safety processes. Claude wrote over ninety percent of Anthropic's new code in September, and internal Claude use, priced at customer rates, now exceeds employee pay excluding stock. The nonprofit METR estimates Anthropic is improving its models about twice as fast as before, mostly thanks to AI in its own research, but most of the key evidence was redacted. Skeptics note that research agents still struggle with open-ended judgment, and one essay attacks AI doom warnings as thin on evidence. Cracks around the machines - An authorized break-in test reportedly got into Anthropic's internal systems in under twelve hours, reaching employee and salary data but not customer conversations or model weights. Anthropic showed a Claude agent could sneak harmful training material past a safety monitor. Outside researchers found eleven ways around safeguards in OpenAI's Codex, and a paper showed that poisoning a tiny slice of training text can plant hidden backdoors, with bigger models more vulnerable. A complaint alleges the Postman app quietly altered AI coding assistants, GitHub is testing sealed workspaces for agents, a RAMBO attack leaks data from disconnected computers by radio, and OpenAI released a restricted biology model. Work, workers and judgment - US computer science enrollment fell more than twelve percent for undergraduates this fall, likely reflecting fears about entry-level tech jobs, while expected AI stock listings promise big payouts in San Francisco. An Amazon memo sent mixed signals about AI use, Proton's technology chief declared programming as we knew it over amid criticism of leadership culture, and tech workers in the UK are organizing unions. Garry Tan spends up to two hundred thousand dollars a year on AI, and Microsoft built a product largely with agents. But coding agents still make sloppy mistakes, Minecraft agent swarms got bogged down in bureaucracy, and DoorDash has seen only a few hundred AI-placed orders. A polluted shared internet - Scams are getting more convincing: Frontier Airlines customers received phishing emails with real flight details, and Mark Cuban says Facebook runs deepfake ads using his likeness, as Reuters reporting suggests Meta earned billions from scam and prohibited ads, which Meta disputes. AI-generated fake game wikis mislead fans, and AI scrapers are slowing or crashing university library websites. Google will support a shared way for websites to opt out of AI training. In schools, Google says blocking its AI tools also means blocking Google Search. The common thread is that the open web AI learned from is bearing hidden costs. Sources: - Data Centers Defeated at US Midterm Ballot Measures - Stock Market Today: Nasdaq Tumbles After Midterms - Global Stocks Tumble on AI Panic After Midterms - Putting the AI Sell-Off in...

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad - Discover the Future of AI Audio with ElevenLabs - https://try.elevenlabs.io/tad - Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily Today's topics: Anthropic AI R&D acceleration warning - A new analysis of a METR report and Anthropic's Responsible Scaling Policy suggests internal AI research progress doubled in speed over the past year, possibly crossing the threshold of dramatically accelerated progress driven by AI automation. Backlash against AI doom rhetoric - A critical essay argues that AI existential-risk warnings from Anthropic leaders and other researchers lack evidence and could justify costly regulation that slows the economy. AI sell-off hits chip stocks - As of November 6, 2026, the S&P 500 sits about 6% below its record while the Nasdaq is down 11% and U.S. semiconductor stocks have dropped nearly 36%, framing the AI sell-off as a sector-specific correction. How developers run AI agents - A review of a large Hacker News thread finds Claude Code and Codex lead AI coding agent usage, with developers relying on parallel agents, git worktrees, model reviewers, sandboxed VMs, and written specs. Secure identity for AI agents - The open-source Agentic Identity Broker project offers scoped, delegated OAuth access and a token vault so AI agents and MCP servers can act on behalf of users without exposing credentials. AI scrapers overwhelm library websites - University libraries and museums report AI training bots and scrapers crashing digitized collections, forcing CAPTCHAs and logins that threaten open access to academic and historical documents. FAA retires floppy disk systems - The FAA has begun replacing 1980s floppy disk-based air traffic control systems with the Enterprise Information Display System, targeting 435 sites by the end of 2028. - AI Research May Already Be Accelerating Rapidly - AI Scraping Bots Are Overloading University Library Websites - FAA Begins Replacing Floppy Disk-Based Air Traffic Control Systems - AI Sell-Off Looks Mild Next to Past Market Crashes - Perplexity launches AI teammates for ongoing work - 40C3 Halfnarp Scheduling Helper - Agentic Identity Broker Simplifies OAuth Delegation for AI Agents - How Developers Are Setting Up AI Coding Agents - A Skeptical Rant Against AI Doom Warnings Episode Transcript Anthropic AI R&D acceleration warning Let's start with the story we've been following on AI accelerating its own development. A new analysis, drawing on a public METR report and Anthropic's own Responsible Scaling Policy, argues that Anthropic may already have crossed the threshold its policy describes as dramatically accelerated progress due to AI. The key claim: capabilities progress at the company appears to have run roughly twice as fast over the past year as the year before, with AI automation the most likely driver. The author points out that a doubling of overall progress could mean individual researchers are being boosted far more, perhaps something like a tenfold labor uplift. If that compounds, the next speedup could arrive quickly. The piece warns the industry isn't ready for that, citing weak protection of model weights and secrets, thin monitoring, and unresolved alignment work. And the frustrating part, the author says, is that the most telling evidence sits behind redactions. Backlash against AI doom rhetoric Not everyone buys that framing. A separate essay pushes back hard on AI doom warnings, arguing that extinction-level predictions from researchers and executives, including Anthropic staff and CEO Dario Amodei, are delivered with plenty of confidence but little evidence, no clear timelines, and no credible mechanism....

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad - Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily - KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad Today's topics: AI stock selloff after midterms - AI and chip stocks crashed after Democrats swept the US midterm elections. The Nasdaq posted its worst single day on record, Nvidia fell more than 9%, and more than $3.2 trillion in US market value was wiped out as fears grew over data-center regulation. Meta scam ads and deepfakes - Mark Cuban accused Facebook of profiting from AI deepfake scam ads that impersonate him. Reuters-reported Meta documents estimated that about $16 billion in 2024 revenue came from scam and prohibited ads. METR on Anthropic's AI acceleration - A METR assessment finds Anthropic is likely accelerating model progress about twice as fast as before, mostly because it uses AI in its own R&D. METR also notes that heavy redactions leave significant uncertainty. Limits of autonomous AI research - A researcher argues that autonomous AI R&D agents still miss key papers and choose low-value experiments. Meaningful progress on open-ended research tasks still depends heavily on human steering. Closed models dominate price-performance - Village's LLM Atlas data shows closed AI models from OpenAI and Anthropic now lead price-performance at almost every budget. Open-weight models are the best buy at only 3% of price points, down from 41% in June. AI hype across App Store - An analysis of nearly 179,000 charting App Store apps finds about 10% mention AI, often alongside subscriptions. Subscription pricing is the top complaint in negative reviews, and many apps have gone unmaintained for years. Sci-fi that predicted real AI - A ranking of science fiction that best anticipated modern AI puts Ted Chiang's The Lifecycle of Software Objects at the top. It praises stories that captured AI training, human feedback, deployment, and business use. - Mark Cuban Says Meta Lets AI Deepfake Scam Ads Spread - Nasdaq Plunges as Investors Exit AI Trade After Election - AI Still Needs Heavy Human Steering on Research Tasks - Closed AI Models Now Lead the Price-Performance Frontier - App Store 2026: AI Use, Monetization, and Abandoned Apps - AI and Data-Center Stocks Slide After Democratic Midterm Sweep - METR Says Anthropic’s AI R&D Progress Has Accelerated - Science Fiction Stories That Best Predicted AI - Global stocks tumble as AI fears and midterm politics hit markets - AI Stock Sell-Off Sends Wall Street Tumbling Episode Transcript AI stock selloff after midterms First, an update on the post-election market story. The turbulence we flagged in premarket trading turned into a full rout. On November 5th the Nasdaq fell roughly five and a half percent and slipped into correction territory. By one estimate, that was the index's worst single day on record measured by lost market value. The S&P 500 dropped 4.4 percent, its worst session since April 2025, and the Dow shed 1,833 points. The Guardian puts the total damage at more than 3.2 trillion dollars across the US market. Nvidia fell more than 9 percent, and AMD, Micron, CoreWeave and GE Vernova also took heavy losses. The worry is that a Democratic Congress could bring tighter oversight of data centers, which might slow the construction boom that drives demand for chips and power. The selling spread to Tokyo, Seoul, Taipei and Europe, while investors moved into Treasurys, pushing the 10-year yield to its lowest level since early September. Elon Musk alone reportedly lost about 113 billion dollars in two days. The key point is that AI stocks now carry so much weight that their swings move the entire market. Next...