Context Window: AI Daily News Brief

Artificially Intimidating
Artificially Intimidating
  • 5.0 (1)
  • News
  • Daily

The 4-minute daily AI news brief that makes artificial intelligence make sense. Every morning, five stories in plain English — no hype, no doom-scrolling, just the signal. artificiallyintimidating.com

  1. 11h ago

    OpenAI Puts ChatGPT and Codex in One Bucket -- AI Brief November 1

    Good day %%first_name%%. OpenAI is about to make ChatGPT and Codex draw from the same usage allowance, DeepSeek published a blueprint for an agent that never goes offline, and four AI labs agreed to honor one robots.txt line that tells them to keep out. Also today: a viral post about an Anthropic talk that doesn’t hold up, and OpenAI raising less money than it wanted. ChatGPT and Codex Will Share One Usage Bucket OpenAI Help Center * What happened: OpenAI says that starting in December, the usage allowance on Plus, Pro and Business plans will be shared across ChatGPT and Codex, its coding agent. Usage will be measured in tokens, and “instant chat will not draw from your usage.” OpenAI hasn’t given a start date or allowance sizes by plan, TokenPost reports. * Why it matters: Chatting and coding used to be two separate habits. Soon a long Codex session will leave less room for everything else you do in ChatGPT, and the reverse. You’ll see your combined remaining usage under Settings → Usage, and at the limit you can wait, upgrade or buy credits. On Friday we covered OpenAI scrapping its ChatGPT Pro cut. This is the next chapter. * What everyone’s saying: The Hacker News thread is running hot, with 106 points and 48 comments when I read it. One commenter says Claude is already ahead on coding and that the merge will push non-coders to switch too. A researcher says the thing he pays OpenAI for is that Pro-tier chat and Codex are billed separately, which suits long math reasoning, and that he’ll “see once this transition actually happens.” * My read between the lines: OpenAI says it expects most users won’t wait for resets more often than they do now. That sentence is carrying the whole announcement, because there are no allowance numbers to check it against. The stated reason is bringing Work’s agent features into the main ChatGPT experience, which makes ChatGPT an agent product with a chat window, billed like one. 📖 Further reading: OpenAI Codex: Apache-2.0, 120,183 Stars, and the Bill That Isn’t in the Repo — Our Codex breakdown: the client is free, and what you can actually run depends on the plan, including an April cut OpenAI announced itself. If your week now includes rationing one shared AI allowance, here is a thought: hand the grunt work to a coworker instead. Viktor is an AI agent that lives in your Slack and plugs into 3,000+ tools, then does the work: reports, dashboards, code and campaigns. Not a chatbot, a coworker. New readers get $50 off their first month. Hire Viktor → DeepSeek Sketches an Agent That Never Logs Off BigGo Finance * What happened: On Saturday evening, DeepSeek’s Harness team published an article describing a “Super-agent”: one assistant that stays online, takes in messages from group chats, GitHub and email, and hands the work to separate worker agents. It also put an MIT-licensed prototype plugin on GitHub. BigGo Finance notes this is an engineering exploration, not a formal DeepSeek product. * Why it matters: The team’s argument is that agents are no longer limited by the model but by you: a person can usually follow only three to five active agent sessions at once. So a lightweight router sorts incoming work, workers keep their own context and tools, and a separate watchdog checks for tasks that stall or loop. DeepSeek says an internal instance already helps reproduce bugs and respond to code reviews. * What everyone’s saying: Coverage so far mostly relays DeepSeek’s own numbers. In its simulated tests, the 95th-percentile task time was 15.7 minutes against 29.5 for a conventional single-session setup, with cache-read tokens up about 37%. The team lists long-term memory, security and privacy, and human oversight as unsolved. * My read between the lines: Look at the test: about 300 simulated messages per run, roughly 20 of them real work requests, five runs each. That shows the plumbing keeps up, not that the fixes are any good. The detail I like is the watchdog. BigGo calls it an AI supervisor, so the system built to save human attention needed a manager of its own. 📖 Further reading: What Is ChatGPT Dots? — OpenAI’s own always-on agent, and why a helpful thing that works while you’re away gets harder to leave every day you use it. A note from me: the daily Brief is free and always will be. If you want the longer, messier version of a story like today’s OpenAI one, members get the paywalled deep-dives plus the full archive. Become a member → Four AI Labs Agree on One Keep-Out Line Engadget * What happened: Google, Anthropic, OpenAI and Microsoft committed on Friday to honor a new robots.txt directive that lets publishers opt out of generative AI training with one rule instead of one per company, Engadget reports. It comes from a proposal at the IETF, the internet standards body, written by Anthropic’s standards lead Kevin Kelley. Sites add a group with the user agent __usage_ai-training, then Disallow rules, with Allow lines for folders like a press kit. * Why it matters: Until now a site had to name each crawler and keep separate rules for every provider. The new token covers training only: it doesn’t block crawling, doesn’t change how a site shows up in search, and can’t undo anything already scraped. Anthropic has started honoring it, Microsoft begins February 1, 2027, Google says early 2027, and OpenAI says “the coming months.” * What everyone’s saying: OpenAI calls it a “voluntary effort” and says it “will not replace the controls we already provide through robots.txt.” Engadget notes Google will read the token as a block on Google-Extended, the crawler that gathers material for Gemini training. One explainer points out that it isn’t an official IETF standard yet. * My read between the lines: A sign on the gate, not a lock. Nothing in the announcement creates a legal duty, and for the next few months the practical effect for a publisher is close to zero. The Google detail is the telling one: the new shared line gets treated as the Google-Extended block publishers could already set. One line of intent is tidier than four rule sets, and that’s the real win. 📖 Further reading: Google’s Invisible Axe: The Silent Killer — Not about training data, but about what happens when Google’s rules apply to you and its support doesn’t answer: Google unlisted our business and every appeal went nowhere. Viral Anthropic Claim Doesn’t Check Out X * What happened: A post from @halvenruns, at about 1,600 likes when I read it, credits an “Anthropic Research Lead” with saying “prompting is basically over,” that “99% of our engineers already run swarms of 300+ self-improving agents,” and that “now everyone is shifting to graphs,” all in a 20-minute talk. A day later the account posted a new version of the same pitch with 95%, no swarms of 300+, a ten-minute video, and an “Anthropic engineer.” * Why it matters: Lines like this get screenshotted into slide decks. Here is what I could find behind it. The Chinese-language account Mai Yang says the 20-minute talk is by Theo Chu, an Anthropic research product manager, at Code w/ Claude Tokyo, covering how Claude’s capabilities are rising, evals, shrinking scaffolding, auto mode and closing the agent loop. The ten-minute “empty terminal” clip, posted to YouTube on October 25, has a transcript that is a Claude Code product manager announcing general availability and the Claude 4 models. I couldn’t find the 99% or the 300-agent figure in either. * What everyone’s saying: The real-world version of the claim is that people are supervising agents, not writing code. Forbes reported on October 26 that Claude now writes 90% of Anthropic’s code. The sharpest reply under the post is shorter: “The guide is still a prompt.” * My read between the lines: If Mai Yang’s summary is right, the advice from the stage was to shrink the scaffolding as models get stronger, which is close to the oppos… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    OpenAI Puts ChatGPT and Codex in One Bucket -- AI Brief November 1

  2. 1d ago

    ChatGPT Trusted Fake Local News -- AI Brief October 31

    Good day %%first_name%%. It’s Halloween, and the scariest thing in today’s Brief is a “local newspaper” with 16 stories, all about one candidate, that ChatGPT cited as a source. Also: Google’s Genie 4 builds worlds with sound that last about two minutes, a Harvard professor says middle managers have more to fear than juniors, and Cal Newport and Tim Ferriss each warn that AI can make work faster and worse at the same time. ChatGPT Cited Fake Local News Sites NPR * What happened: NPR ran 37 questions about 10 competitive U.S. House races through ChatGPT and Gemini. In 19 of ChatGPT’s answers, about half, it cited at least one site from a network of look-alike “local news” outlets, and in 17 of those it was an outlet posing as local news. One of them, the Kalamazoo Beacon, has published 16 stories, all about one Democratic candidate, and ChatGPT put his affordability plan at the top of its answer. * Why it matters: The sites are thinly veiled political ads. The security firm VisionHeight counted 37 of them, all set up on September 15, all but one within an eight-minute window, and no one has identified the operator. They welcome AI crawlers, publish in question-and-answer form, and ship llms.txt files, which most sites don’t. Chatbot answers are now an advertising target. * What everyone’s saying: Georgetown’s Renée DiResta told NPR that disclosing who is behind the sites would be “obviously the ethical thing to do.” OpenAI said ChatGPT’s search tools “cite from a range of sources so people can explore and make their own judgements.” Gemini cited none of the sites, and Google said it took “swift action” after detecting a policy violation. * My read between the lines: Nobody hacked anything. The sites look like exactly what a chatbot is built to trust: local, newsy, question-shaped, and up against thin competition, a gap researchers call a data void. The telling detail is that the Kalamazoo Beacon was still cited every time in five re-runs after NPR flagged it to OpenAI. NPR ran each query once, so treat the 19 as a snapshot. The playbook works for any party, which should worry everyone. 📖 Further reading: AI Agents Can’t Find Your Business. I Tested Mine. — The flip side of today’s story: what an AI agent can and can’t read about a business, tested on a real one. Tim Ferriss says AI multiplies whatever you hand it, so hand it work worth multiplying. Viktor is an AI agent that lives in your Slack and plugs into 3,000+ tools, then does the work: reports, dashboards, code and campaigns. Think coworker, not chatbot. New readers get $50 off their first month. Hire Viktor → Google’s Genie 4 Adds Sound to Its Worlds Ars Technica * What happened: Google DeepMind announced Genie 4 on October 27, a research preview of a world model that generates interactive scenes with native audio and more realistic physics and people, Ars Technica reports. You steer it with text mid-simulation, like changing the weather or moving the camera, and Gemini 3.8 Flash acts as a guiding agent. It is in the hands of trusted testers now, per DeepMind, with Google AI Ultra subscribers next. * Why it matters: DeepMind’s pitch is endless training environments for robots and agents. In one demo, Genie 4 and Gemini 4 Argon simulated Apptronik’s Apollo humanoid playing checkers. Earlier this week we covered Google handing Gemini 4 Carbon to cyber defenders first. This is the other side of the Gemini family, the one that builds pretend worlds. * What everyone’s saying: TechCrunch frames it as a crowded race. It arrives a couple of months after Runway’s second world model, while Fei-Fei Li’s World Labs is being acquired by AMD, Amazon is in talks to buy Decart, and Rivian just bought Third Dimension. * My read between the lines: The caveat is the part to remember: DeepMind says Genie 4 worlds stay consistent for only a couple of minutes before the model forgets how things started. A robot that rehearses in a world with a two-minute memory is learning from a dream that keeps changing its mind. What could possibly go wrong? Mostly everything after minute two. The daily Brief is free and always will be. If you want the story behind the headline, members unlock the paywalled deep-dives plus the full archive. Become a member → Harvard Professor Flips the AI Job-Risk Script Silicon Valley Girl * What happened: Harvard Business School professor Karim Lakhani told the Silicon Valley Girl podcast that he doesn’t think junior work is being automated. The people more at risk, he says, sit in middle management, because inside companies the pushback against AI usage comes from the 35-to-55 crowd while juniors run circles around the organization. * Why it matters: His case rests on how people use AI, not whether they do. In his study of 244 BCG consultants, the interviewer cited 27% as “self-automators” who delegated everything. The paper says that group built neither domain expertise nor AI skills, while “centaurs” deepened their expertise and “cyborgs” picked up new AI skills. * What everyone’s saying: Lakhani says he has fired people who told him “the AI gave me that,” and that if all you do is copy and paste between a person and the model, “that job is gone.” He asks law-firm and manufacturing CEOs the same question: if you are a thin wrapper around a model, why should you exist? * My read between the lines: He is careful to say the question isn’t settled. Colleagues he respects worry about the junior side, and he notes the soft entry-level job market has many causes, not just AI. It is an observation from inside companies, not a labor-market study. The finding that sticks is the self-automator one: you can use AI every day and learn nothing. Newport: AI Agents Are a Cognitive Assembly Line Cal Newport * What happened: In a new video, Cal Newport takes a Jerry Seinfeld clip he heard in the Acquired podcast’s Seinfeld episode and applies it to AI agents. Asked why he didn’t find a more efficient way to make his show, Seinfeld answered: “If you’re efficient, you’re doing it the wrong way. The right way is the hard way.” * Why it matters: Newport’s thesis is that technologies redefine what work means. Email, he argues, was pitched as a better fax and voicemail and ended up making work a constant back-and-forth of messages. Agents, he says, are pushing a new definition: knowledge work as a “cognitive assembly line” of small tasks, like updating the deck and sending the email, that AI can speed up. * What everyone’s saying: The agent pitch writes itself: more tasks per day feels like productivity. Newport’s counter is that this version of the job has you supervising agents all day, which he calls “a pretty miserable definition of work,” like being a manager without any human interaction. * My read between the lines: His fix is unglamorous: decide what good work looks like in your field before a tool decides for you. Seinfeld’s point is that the valuable work takes weeks, and whether it takes 19 days or 21 barely matters. That makes a handy test for any agent pitch: does it make the valuable thing better, or just the clerical thing faster? Tim Ferriss: AI Multiplies the Useless Tim Ferriss * What happened: In a clip from a Q&A posted Friday, Tim Ferriss said the first two steps of his DEAL framework from The 4-Hour Workweek, definition and elimination, matter more in a world of agents. His reasoning: AI is an accelerant, and if you haven’t identified the critical few things first, it “simply multiplies the useless.” * Why it matters: Ferriss says that in more than 90% of the AI use cases he has seen, people take things that probably shouldn’t be done at all and do them very fast, creating more work, not less. He calls the idea that AI will create infinite leisure “a completely false promise” and expects AI used as a blunt instrument to bring more work and more stress. * What everyone’s saying: The popular story is that agents tak… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    ChatGPT Trusted Fake Local News -- AI Brief October 31

  3. 2d ago

    Anthropic Is SpaceX’s Biggest Customer, For Now -- AI Brief October 30

    Good day %%first_name%%. Anthropic turns out to be SpaceX’s biggest customer, and it can leave on 90 days’ notice. OpenAI backed off a price cut, a bank runs its own coding model, and Google and Microsoft are splitting on how to charge for agents. But first, a quick plug for the new podcast episode below. New on the podcast, episode 3: novelist Justin Feinstein wrote Your Behavior Will Be Monitored, a novel about an ad guy hired to teach a bot how to market to people inside a company where the bots watch the employees. He figured that world was five to ten years away. Then it started happening before the book was out. We get into the cookie data that knows too much (including the custom Oreo ads that showed up 15 minutes after an email), why he drives differently around Waymos, how to spot AI writing, and the wave of AI-powered crime he is writing about next. Listen to episode 3 → Follow the show in any app: pod.link/6785740580. Anthropic Is SpaceX’s Biggest Customer, For Now Bloomberg * What happened: Bloomberg reports that “Customer B,” the unnamed client behind about 34.5% of SpaceX’s third-quarter revenue, is Anthropic, citing a person familiar with the matter. The money flows to SpaceX’s AI division, now called Super Intelligence, which supplies computing power. Bloomberg’s headline puts the deal at $3 billion a month. * Why it matters: Anthropic’s IPO filing, which Yahoo Finance covered, says it agreed to spend up to $120.5 billion with SpaceX’s AI arm over terms into 2029, about $3.1 billion a month once all the capacity arrives in December. Once the initial non-cancelable periods end in early 2027, either side can cancel with 90 days’ notice, Gizmodo notes. Earlier this week we covered Anthropic opening its books. This is the next chapter. * What everyone’s saying: Yahoo Finance reads the deal as proof that SpaceX’s AI compute business is durable. Gizmodo’s angle is the size: the filing shows the SpaceX agreement is tens of billions of dollars bigger than the $84.5 billion earlier reports had suggested. * My read between the lines: “Durable” and “cancel on 90 days’ notice” are doing a lot of work in the same story. SpaceX’s biggest customer can leave in a quarter, and the landlord also builds Grok, a rival to Claude. Colossus is a very expensive apartment building with one big tenant and a very short notice clause. Microsoft wants to put a meter on your AI, and your Friday still has a to-do list. Viktor is an AI agent that lives in your Slack and plugs into 3,000+ tools, then does the work: reports, dashboards, code and campaigns. Think coworker, not chatbot. New readers get $50 off their first month. Hire Viktor → OpenAI Scraps Its ChatGPT Pro Cut The Decoder * What happened: OpenAI doubled the Work and Codex usage quotas on its Pro 100 and Pro 500 plans. On Pro 200 it doubled the lower quota and reversed a planned cut after user pushback, The Decoder reports. The Plus plan stays the same, and new and existing subscribers get the new quotas automatically. * Why it matters: On these plans the usage allowance is the real price. The $200 sticker never moved, but what you got for it changed, and then changed back. OpenAI’s own caveat: doubling a quota does not mean twice as many finished tasks, because consumption varies by model, task type and settings. * What everyone’s saying: The Decoder frames it as a race with Anthropic, which is preparing its IPO, with both companies competing hard for business customers. OpenAI says it will keep shipping efficiency gains and points to its new 6.1 models, Terra and Sol, which The Decoder says are much more efficient while matching or beating their predecessors. * My read between the lines: When a plan sells by the quota, cutting the quota is cutting the product, and customers noticed in about a month. Watch the wording too. A bigger allowance is a promise about a number, not about your Tuesday. 📖 Further reading: Claude Is Burning Through Your Limit Faster Than Ever. Anthropic Won’t Tell You Why. — Allowances are the battleground now, and this one covers how usage limits get eaten and how to stretch yours. The daily Brief is free and always will be. If meters, ads and who pays for the agents have you wanting the how-to behind the headlines, members unlock the paywalled deep-dives plus the full archive. Become a member → Revolut Runs Its Own Coding Model Revolut’s David Abitbol * What happened: David Abitbol, who leads AI operations at Revolut, a digital bank with 80 million customers, says in a conference talk that the company self-hosts the open GLM coding model for its 2,000 engineers on 56 Nvidia B200 chips. The talk’s title claims the move cut AI coding costs by 90%. * Why it matters: Revolut started with strict budget limits after seeing engineers spend thousands of dollars a month with little return. Self-hosting made the billing flat, with no month-end bill shock, so engineers can use tokens guilt-free. Abitbol says GLM 5.1 showed a 70% improvement in lead time, versus 10% from Google’s Gemini when it first launched. * What everyone’s saying: The engineering is the real story. Abitbol’s target is the first token in under three seconds, because past five his Slack starts to blow up. On the same 56 GPUs, Revolut pushed demand from about three requests a second in April to over seven by August, while time to first token fell from more than 10 seconds to around 1.6. * My read between the lines: Open models are winning on the bill, not the benchmark, and the frontier models stay at Revolut as the exception for the hardest work. But “free” tokens have rent. Revolut pays it in 56 of Nvidia’s newest chips and a team that can tune them. Google and Microsoft Split on Paying for Agents Benzinga * What happened: On Wednesday’s earnings calls, Microsoft CEO Satya Nadella said Copilot’s business model is “expanding from per-seat subscriptions to include consumption,” and that Copilot customers using usage-based credits grew over 75% in the quarter, Benzinga reports. Alphabet CEO Sundar Pichai said Google will bring its enterprise agents to consumers later this year, starting with the Gemini app’s 1 billion users. * Why it matters: Both companies are selling software that does work for you, with opposite cash registers. Microsoft meters the work. Google’s chief business officer, Philipp Schindler, said he is “really excited about the advertising opportunities in an agentic world.” In plain terms: one company bills you when the agent works, the other may also get paid by whoever wants the agent to pick them. * What everyone’s saying: The market picked a side on the day. Microsoft’s $92.28 billion in revenue, up 19%, beat estimates. Alphabet’s $127.58 billion missed by about $620 million, and its shares fell nearly 10% after hours, per Benzinga. * My read between the lines: An agent with advertisers is a concierge on commission. Meters are easy to love when you are the one reading them, but an assistant that books your flight is only useful if it is not also taking bids. 📖 Further reading: AI Agents Can’t Find Your Business. I Tested Mine. — If agents become the front door for shopping and booking, this is the test for whether they can even find you. Y Combinator Pits Claude Swarms Against GPT Swarms Y Combinator Labs * What happened: Y Combinator’s research arm, working with Olam Labs, ran swarms of AI agents in Minecraft and in Mindustry, a factory-building strategy game, in a new write-up. In Mindustry, teams of GPT-6.1 Sol agents in Codex beat teams of Claude Opus 5.5 agents in Claude Code, 6 games to 1, over about 9.5 hours. * Why it matters: A swarm is a group of agents working toward one shared goal, and how they organize is an AI safety question as much as a game one. In the Minecraft test, 50 GPT agents had three hours to “build a thriving civilization.” They spent it rationing food … This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    Anthropic Is SpaceX’s Biggest Customer, For Now -- AI Brief October 30

  4. 3d ago

    Google Hands Its Strongest Model to Cyber Defenders First -- AI Brief October 29

    Good day %%first_name%%. Google handed its new Gemini model to a few cyber defenders before anyone else, and Apple finally let you touch a Mac. Also today: GitHub’s agents now start 40% of pull requests, Bloomberg gives quants a coding assistant that asks permission first, and an Anthropic manager retires the coding interview. Google Gives Gemini 4 Carbon to Defenders First 9to5Google * What happened: Google announced Gemini 4 Carbon on Wednesday, October 28, two weeks after it launched Argon, calling it its next frontier model with “significant gains across coding and agentic tasks,” 9to5Google reports. It goes to a subset of cyber defenders first, with developers, businesses and consumers to follow. * Why it matters: The model is good at the offensive side of security too. Google’s numbers: 88.2% on a vulnerability-finding test, 54.8% on exploitation and 71.0% on patching. So the first access program, called Fairwind, comes with conditions, per Unite.AI: security teams only, phishing-resistant multi-factor login, and no reselling. * What everyone’s saying: Google’s pitch is leaderboard-shaped: first on the Artificial Analysis Intelligence Index at 59, and up 89 Elo points to fourth on WebDev Arena. Intro pricing is $2 per million input tokens and $10 per million output, rising to $4 and $20 later, 9to5Google says. * My read between the lines: Staged access for the scary skill is becoming the industry move. Anthropic did it in April, and Google says it is also working with the US government’s voluntary pre-release testing process. Note the price: the intro rate doubles, so the teaser is the rate you get while the model is still a rumor. 📖 Further reading: Anthropic built the most powerful AI ever. You can’t use it. — Handing a cyber-capable model to defenders first is the pattern Anthropic set with Mythos and Project Glasswing, and I wrote up what the access gap means for the rest of us. GitHub says agents now start 40% of pull requests, so your Thursday could use a few of its own. Viktor is an AI agent that lives in your Slack and plugs into 3,000+ tools, then does the work: reports, dashboards, code and campaigns. Think coworker, not chatbot. New readers get $50 off their first month. Hire Viktor → Apple Puts a Touchscreen on the Mac CNBC * What happened: Apple announced the MacBook Pro with ProLED XDR Display, its first Mac with a touchscreen, powered by M5 Pro or M5 Max chips, CNBC reports. The 14-inch starts at $2,899, the 16-inch at $3,399, and it ships November 6. “For the very first time,” said CEO John Ternus, “we are bringing touch to the Mac.” * Why it matters: Apple spent years refusing touch on laptops. The AI angle is the memory. 9to5Mac lists up to 128GB of unified memory, and CNBC says Apple is pitching its chips as the best place to run AI on the device instead of in the cloud. In May we covered The Mac Mini Sold Out Because of You, and this is the sequel. * What everyone’s saying: CNBC says Apple teamed up with Meta to show large models running on clusters of Macs wired together at high speed, so developers can run agentic code without paying Anthropic or OpenAI. Meta’s model team bought hundreds of Mac Minis this year to test agents. IDC’s Ryan Reith was less excited about touch itself: Windows laptops have had it for ages, he said, and it is “just another feature, not a selling point.” * My read between the lines: The touchscreen is the headline and the cluster is the story. A laptop line that scales to 128GB of memory, plus a way to chain Macs together, is Apple’s offer to the people who now pay by the token. Reith’s line is the tell: he says the draw is developers, not shoppers who want to poke a screen. 📖 Further reading: Your Mac Just Became a $20/Month AI Employee — If the Mac is where agents are headed, this is the guide to putting an AI employee to work on one. The daily Brief is free and always will be. If keys, agents and interview rules have you wanting the how-to behind the headlines, members unlock the paywalled deep-dives plus the full archive. Become a member → GitHub: Agents Now Start 40% of PRs The Next Web * What happened: At GitHub Universe on Wednesday, GitHub launched Copilot Fusion, a system that picks which AI model, or mix of models, handles each step of a coding task, The Next Web reports. By September, more than 40% of pull requests began with a draft from an agent, up from 17% in March. * Why it matters: A pull request is a proposed code change that a person reviews before it goes live. Agents now write the first draft of a big share of them. DevOps.com says GitHub also made Copilot Projects generally available, a tool for tracking many agents working in parallel, plus a deeper code review option called Max. * What everyone’s saying: Microsoft CEO Satya Nadella posted that “we’re redefining the SDLC for the agentic era,” and said Fusion delivers frontier-level quality 40% faster at up to 77% lower cost. Enterprise customers can now use every Copilot feature without a separate subscription, paying for the AI credits they use instead, per The Next Web. * My read between the lines: Billing by usage instead of by seat puts a meter on the work, which sits oddly beside yesterday’s Amazon guidance calling token counts a flawed scoreboard. Also, “started by an agent” is not “merged.” The reviewer at the end of that conveyor is still a person, and DevOps.com says GitHub is moving to a multi-region Azure setup by early 2027 to keep up with all the traffic. Bloomberg’s Coding Assistant Asks Before It Runs Bloomberg * What happened: Bloomberg introduced BQuant Assistant, a conversational AI coding assistant built into BQuant, its Python environment for finance, Bloomberg says. You type a question in plain language and it writes Python and queries in Bloomberg’s own query language, BQL, for data across equities, fixed income, FX, commodities and more than 22,000 fields. * Why it matters: It does not just suggest code. Subject to your approval, it inserts the code into your project and runs it in BQuant’s secure sandbox, and it asks before it reads from or writes to a notebook. It is generally available in BQuant Enterprise, with a limited beta in BQuant Desktop at the end of October. * What everyone’s saying: Bloomberg CTO Shawn Edwards framed it as speed: investors have bigger, messier datasets, and turning an idea into repeatable analysis still takes specialized coding skill. The company says it helps everyone from a portfolio manager sketching a first query to a quant building a multi-asset signal. * My read between the lines: The approval step is the product. In finance, the assistant may write the query but a human presses the button, which is the same trade Claude Code just loosened for deleting files. The release also never names which AI model is underneath. The pitch is Bloomberg’s data, not anyone’s brain, and that is a choice. Anthropic Manager Retires the Coding Interview The Next Web * What happened: Felix Rieseberg, an Anthropic engineering manager who has led engineering for Claude.ai, Claude Code Desktop and Cowork, wrote on Tuesday, October 27 that he is dropping coding interviews for his team, The Next Web reports. His comparison: there is no more point in giving someone a little React challenge than in quizzing them about assembly, because Claude is now better at both. * Why it matters: He says those small challenges were a weak but useful predictor of success, and his most impactful builders now succeed whatever they score on them. “Hunger beats tenure,” he wrote, noting newer engineers are flying past people who are probably better coders. Yesterday’s Brief had Goldman linking AI use to weaker hiring; this is the hiring side of the same shift. * What everyone’s saying: The Next Web places it in a wider scramble. Google is expanding p… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    Google Hands Its Strongest Model to Cyber Defenders First -- AI Brief October 29

  5. 4d ago

    Claude Code Skips the Parachute and Trump Wants ‘AI’ Banned -- AI Brief October 28

    Good day %%first_name%%. Anthropic gave Claude Code a switch for deleting files without asking, and the President wants to ban the word “AI.” Also today: Amazon tells its engineers to stop counting tokens, Goldman’s careful note on AI and hiring, and Airbnb’s AI bouncer for Halloween weekend. Claude Code Adds a Switch for Deleting Anything The Register * What happened: Anthropic’s latest Claude Code build, v2.1.311, supports a new environment variable, CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS, that lets the agent run the file-deleting rm command without asking, The Register reports. Anthropic’s Daisy Hollman announced it, noting that the existing skip-permissions flag “doesn’t actually approve EVERYTHING,” because Claude Code still blocks “some obvious things like rm -rf ~.” * Why it matters: The Register says rm and rmdir were among the few actions that still triggered a permission prompt even in bypassPermissions mode. This variable removes that last restraint. Back in April we covered An AI Agent Deleted a Startup in 9 Seconds, and The Register notes there are “numerous accounts” of agents issuing commands like rm -rf ~ that wipe drives. * What everyone’s saying: Some people did not believe it, partly because of the variable’s name and partly because the change is missing from the v2.1.311 changelog, per The Register. Hollman replied, “Not a joke! I hope this helps, and I hope the variable name is memorable.” The Register’s own verdict was drier: “you only live once.” * My read between the lines: The name is the guardrail. Nobody sets a variable called EVEN_MORE_DANGEROUSLY by accident, so the label does the work a confirmation dialog used to do. Context matters too: The Register says auto mode, which runs background safety checks, became the default in August, so this switch is for people who found even that too many checks. And it notes managed settings may still limit it, which is where company IT teams come in. Claude Code now has a switch for skipping its last permission prompt. Your Wednesday has plenty of prompts worth skipping too, the harmless kind. Viktor is an AI agent that lives in your Slack and connects to 3,000+ tools, then does the work: reports, dashboards, code and campaigns. Not a chatbot. A coworker. New readers get $50 off their first month. Hire Viktor → Amazon Tells Engineers to Stop Tokenmaxxing CNBC * What happened: Amazon posted guidance on its website Tuesday, October 27, that calls token usage, AI-tool adoption rates, and code changes or reviews “deeply flawed” ways to measure how teams use AI, CNBC reports. The memo went around internally in September. “Tokenmaxxing” means pushing AI usage as high as possible, whether or not it produces anything. * Why it matters: CNBC says companies are grappling with AI’s surging costs and with how to tell whether the technology pays off. Amazon says those metrics can lead to higher costs and too much complexity. Yesterday’s Brief had Bill Gates floating a token tax; today a company is telling its own staff that counting tokens is a bad scoreboard. * What everyone’s saying: The memo is more balanced than the headline. It stops short of telling staff to ignore the numbers, saying token usage and code generation should be weighed “with care and judgment.” It prefers measuring “long-term outcomes,” and it says that “If your work does not benefit from current AI tooling, that is a finding, not a failing.” It also insists that “AI will be at the center of software engineering practice.” * My read between the lines: This is Goodhart’s law with a company letterhead: when a measure becomes a target, people hit the target. “A finding, not a failing” is the sentence to frame, because it gives an engineer permission to report that the tool did not help, which is hard to do while adoption numbers are everywhere. The Amazon post and the push to use AI sit in the same document. Read the push as the policy and the metrics warning as the footnote. 📖 Further reading: I found 350,000 tokens hiding in plain sight — Amazon says a big token count tells you little about the work; this is what I found when I audited my own scheduled tasks. The daily Brief is free and always will be. If a week of rules about tokens, names and agents has you wanting the how-to behind the headlines, members unlock the paywalled deep-dives plus the full archive. Become a member → Trump Vows to Ban the Term “AI” Variety * What happened: In a Tuesday Truth Social post, President Trump asked when The Wall Street Journal would start calling AI “SI (Super Intelligence),” and wrote, “In short order, we will be officially BANNING the term, ‘AI/Artificial Intelligence,’ in all relevant documents, and everything else, due to its inaccuracy.” Variety has the full post. * Why it matters: This follows his September 29 executive order, which told the executive branch to use “Super Intelligence” and “SI” and to “not acknowledge the usage” of “AI” in any applicable setting, per Variety. We covered the first push in our September 26 Brief. Deadline notes the president’s authority to force public adoption is limited. * What everyone’s saying: Almost nobody has switched. Variety says most AI companies, including Anthropic and OpenAI, still call it AI. The National reports several US agencies promised to use SI, but most tech companies, experts, media outlets and governments abroad kept the old term, and Nvidia’s Jensen Huang kept correcting himself at a White House event in early October. * My read between the lines: The National ties the push to polls showing Americans increasingly negative about AI over job losses and data centers. A rebrand is a way to move the label when you cannot move the number, and “super” is a stronger word than “artificial” for people already nervous. Deadline also points out that the White House released an “AI Action Plan” last year and cites the Gulf of America fight with the AP, where a judge said the White House was likely violating the First Amendment. That ruling is on appeal. Goldman Ties AI Use to Weaker Hiring, Not Proof TokenPost * What happened: A Goldman Sachs client note dated October 26 found statistically significant relationships between regular business use of AI and weaker employment growth and higher unemployment across US nonmanufacturing industries, TokenPost reports. It compared the latest three-month averages with 2015 to 2019, and it did not establish that AI caused either outcome. * Why it matters: Four days earlier, per Fortune, Goldman economist Joseph Briggs wrote that up to 9% of workers could be “displaced during the AI transition,” with slower hiring weighing on employment by another 3%, roughly one in 11 workers, if AI eventually lifts productivity by 15%. Same note: concerns about imminent displacement “have cooled in recent months after hitting a fever pitch this spring.” * What everyone’s saying: Goldman’s own framing is calm. Labor markets across developed economies remain solid overall, and its base case is that AI investment eventually raises productivity and profits while disruption may weigh on activity in the near term, per TokenPost. Fortune’s headline put it as “the layoffs are likely waiting for a recession.” Back in August we covered a study showing AI layoffs eating the productivity. * My read between the lines: Two notes in one week, and both can be true: nobody is mass-firing, but where AI is in regular use, companies seem to be hiring less. TokenPost’s own caveat is the part to keep: the measures moved together, which does not show AI caused it. Industries already cooling could also be the ones adopting. A September Goldman analysis found US call-center employment 39% below trend, so the hiring gap has a face already. Airbnb Points Its AI at Halloween Parties News 12 New Jersey * What happened: With Halloween weekend close, Airbnb says it is using artificial intelli… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    Claude Code Skips the Parachute and Trump Wants ‘AI’ Banned -- AI Brief October 28

  6. 5d ago

    Anthropic Opened Its Books and OpenAI Opened a Coupon Aisle -- AI Brief October 27

    Good day %%first_name%%. Anthropic opened its books, and the first thing in them is an $80.7 billion loss. OpenAI, meanwhile, opened a coupon aisle. Also today: Bill Gates puts a name on the AI tax, Grok’s seven-minute Odyssey, and a Chinese open-weight model within three points of the frontier. Anthropic Opens Its Books: $16.3B In, $80.7B Lost Reuters * What happened: Anthropic made its IPO filing public on Monday, October 26, planning a Nasdaq listing under the ticker “ANT.” Reuters reports first-half 2026 revenue of about $16.3 billion against a net loss of $80.7 billion, largely from a rise in the fair value of bonds convertible into shares. CNBC puts $72.4 billion of that loss down to non-cash accounting rules on convertible debt. * Why it matters: It is the first audited look inside the company behind Claude. CNBC says revenue rose 1,304% from $1.16 billion a year earlier, and Reuters expects the listing to set the benchmark for how Wall Street values labs like OpenAI. Reuters also notes some investors now value Anthropic near $1.5 trillion, down from about $2 trillion a week earlier, with an investor roadshow and offering expected in mid-to-late November. * What everyone’s saying: Cautious optimism. DA Davidson’s Gil Luria told CNBC the net losses are “very small” once the accounting charge is set aside, and that the filing undercuts the claim that Anthropic “destroys value with every … AI query.” Freedom Capital’s Paul Meeks was “not wowed” but “modestly pleasantly surprised,” and CNBC flags a substantial slowdown in revenue growth from the second to the third quarter. * My read between the lines: The $80.7 billion is mostly a scoreboard quirk: the convertible bonds got more valuable, so the company booked the rise as a loss. The line to watch is the blank one. The S-1 leaves “run-rate revenue,” reported in August at around $65 billion, empty and says it “is not a substitute for revenue.” And seven co-founders will hold 50.1% of the voting power through Founder LLC, so public buyers get a stake in a public benefit corporation, not a steering wheel. 📖 Further reading: Everyone’s panicking about the Anthropic IPO. They’re wrong. — my April take, written when the IPO was still a Bloomberg rumor: the risk was never the ticker symbol, it is how much of your workflow lives inside one company’s context. Anthropic’s filing runs to more than 90 pages of risk factors. Your Tuesday probably has a smaller pile: the report nobody wants to assemble. Viktor is an AI agent that lives in your Slack and connects to 3,000+ tools, then does the work: reports, dashboards, code and campaigns. Not a chatbot. A coworker. New readers get $50 off their first month. Hire Viktor → OpenAI Is Coupon-Clipping Anthropic’s Customers The Next Web * What happened: The Information reported Saturday that OpenAI is discounting hard to pull enterprise customers away from Anthropic. TNW, citing the report, says Replit moved most of its AI spending to OpenAI after an offer of more than 30% off a contract worth over $100 million a year, and that Anthropic would not extend its earlier discount without a bigger spending commitment. * Why it matters: The Decoder says Rippling’s coding work now runs 52% on OpenAI models, up from 16% in May, while Anthropic’s share fell from 57% to 18%. DataCamp’s CEO said OpenAI’s pricing cut his costs by 80 to 90 percent, and a US bank with more than $7 billion in assets had its ChatGPT seats doubled to 500 at no extra charge, paying under $7 per user instead of the usual $20 or more. * What everyone’s saying: Replit’s Amjad Masad said at TechCrunch Disrupt it was “a decision based on both cost, partnership, and performance,” then added that “Replit remains model agnostic.” The Decoder says OpenAI is partly buying market share with its own margins, though more efficient models have cut its real costs too. TNW notes it has not independently verified the figures. * My read between the lines: This landed the same weekend Anthropic’s filing went public, and The Decoder says Anthropic’s growth slowed from the second quarter to the third, per an investor memo. TNW adds that Anthropic ends discounts once customers use up committed tokens. OpenAI’s pitch is simple: we will be the cheap one while we are the ones who need the customers. A bank getting free seats is a customer-acquisition budget wearing a price list’s clothes. The daily Brief is free and always will be. If you would rather get ahead of the price war than read about it, members unlock the paywalled deep-dives behind the headlines, plus the full archive. Become a member → Bill Gates Backs a Token Tax, Warns of 20 Rough Years The Observer * What happened: In an exclusive interview with The Observer, published October 25, Bill Gates said the business incentive is “hire the robot, fire the human and don’t look back.” The paper reports he wants taxes on AI use, perhaps a “token tax,” and that he thinks billionaires should pay more. * Why it matters: On The Verge’s Decoder podcast, per Business Insider, Gates said the roughly 20-year road to “true abundance” will be “as tumultuous a period as there’s been in human history,” with telesales, telesupport and accounting going first and humanoid robots later taking blue-collar work. He told The Observer that governments will need new revenue for people put out of work, because “the revenue for the safety net to help us through this transition will actually be reduced, even though overall prosperity gets better.” * What everyone’s saying: Business Insider lines Gates up with a chorus: Musk has proposed government checks for all, Anthropic’s Dario Amodei has warned of an “unprecedented” short-term shock, and Bezos and Dimon have both floated shorter workweeks. Sam Altman said in May that he no longer expects a “jobs apocalypse,” and BI notes national unemployment has not risen materially during the AI boom. * My read between the lines: Read the tax math, not the doom. Gates says he has paid $14 billion in tax over his lifetime and would have been willing to pay about $40 billion, “a little more than double.” He is describing a budget problem more than a jobs problem: the money that funds the safety net shrinks just when it is needed. And a “token tax” is a tidy name for taxing the one thing every AI product already meters. Grok Delivered Seven Minutes of “The Odyssey” TechCrunch * What happened: Three months after Elon Musk declared that Grok would release a full-length, “historically accurate” movie of “The Odyssey,” Grok Imagine posted seven minutes on X on Friday, October 23, which TechCrunch says the company is now calling “Episode 1 of our Odyssey series.” It was made with Wonder Studios and director Dev Walker. * Why it matters: It is a live test of whether AI video can carry a long story. The project began as a jab at Christopher Nolan’s “The Odyssey,” which TechCrunch says became a box-office success after its July release. Yesterday’s Brief covered another Grok test, a vote study; this one tests taste. * What everyone’s saying: TechCrunch’s Anthony Ha was not kind. He says a few shots could pass for real actors on a real sea, but the whole feels “completely airless,” with “no reason to care about anyone or anything on-screen.” The backstory voiceover “seems to go on forever,” and he was “listlessly checking” how much time was left. * My read between the lines: Look at the ask in Grok’s own post: it invites viewers to say whether they would like “a full breakdown of how it was done.” That is an offer to talk about the tool, not the film. A promised full-length movie became “Episode 1,” and the proof of the technology is a clip critics describe as lacking momentum. The demo is working; the drama is not. Kimi K3.1 Pro Narrows the Open-Weight Gap to Three Points CGTN * What happened: Moonshot AI released Kimi K3.1 Pro on Friday, and CGTN reports it debuted fourth on the Artificial Ana… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    Anthropic Opened Its Books and OpenAI Opened a Coupon Aisle -- AI Brief October 27

  7. 6d ago

    AI Has a Graveyard Now, and Washington Wants AI CEOs Quiet -- AI Brief October 26

    Good day %%first_name%%. Washington has a message for AI executives, and it is not “thank you.” Also today: TechCrunch’s graveyard of dead AI products, banks finding AI agents inside customer accounts, a vote test that split Grok and GPT, and Arize’s pitch for what an AI engineer does now. TechCrunch Opens an AI Graveyard TechCrunch * What happened: TechCrunch published a running list of AI products and startups that shut down, pivoted or missed badly. OpenAI has several entries: Sora, shut down in March 2026; ChatGPT Atlas, the stand-alone browser, discontinued August 9 after less than a year; Operator, folded into ChatGPT; and a July 9 “super app” redesign with separate Chat, Codex and Work modes that was rolled back soon after users found it confusing. * Why it matters: TechCrunch cites S&P Global Market Intelligence estimating that about 42% of AI initiatives are ultimately abandoned by their corporate parents, for reasons from funding and scaling to weak demand. The startup casualties share a pattern: Relay, a five-year-old Zapier alternative, closed in August once OpenAI, Google and others built similar automation into their own tools, and Notion Mail is closing because users were handing their inboxes to separate AI agents instead. * What everyone’s saying: TechCrunch’s own framing is that the list is “less a scorecard of flops than a record of what these attempts can (hopefully) teach the rest of the industry.” It also notes that OpenAI folding stand-alone apps into ChatGPT has “worked out better than the redesign did.” Character.ai’s bankruptcy filing, five years after launch, sits near the top. * My read between the lines: Read the list for what died and who killed it. The pattern is features that were never products: Relay, Huxe, Notion Mail all lost to a bigger platform adding the same thing for free. Even OpenAI’s own deaths are mostly relocations, with Atlas and Operator moving house into ChatGPT. In this graveyard, a lot of the headstones mark a merger. 📖 Further reading: Thanks to Apple, Your Favorite AI Tool Is a Dead Tool Walking — my June piece on Apple renting Google’s Gemini for the new Siri after the $250 million settlement over the Siri that didn’t ship, the same Siri entry on this list, seen from the platform’s side. Washington’s new message to AI executives is “do less.” Your Monday could use the opposite from one teammate. Viktor is an AI agent that lives in your Slack and connects to 3,000+ tools, then does the work: reports, dashboards, code and campaigns. Not a chatbot. A coworker. New readers get $50 off their first month. Hire Viktor → Washington to AI Executives: Please Stop Talking Axios * What happened: Axios reports that lawmakers, Hill staffers and Washington insiders in both parties wish AI founders would do what other industry chiefs do and just build their products, safely. They say the message from AI executives too often swings between prosperity and doom, sometimes in the same appearance, feeding a strong public backlash. Sen. John Curtis (R-Utah) said the industry is “tripping over their own shoelaces in trying to sell the value of this to the public.” * Why it matters: Administration officials are again frustrated with Anthropic. Axios says Dario Amodei attended a late-September gathering with President Trump without mentioning safety incidents in which Anthropic models affected government websites; the company told officials on Thursday, October 8, then published a Friday-evening blog post. “They were hiding the ball,” one official said. Axios also says both Anthropic and OpenAI are targets of a developing FTC investigation. * What everyone’s saying: Rep. Ro Khanna (D-Calif.) said of OpenAI’s Sam Altman, “You cannot trust billionaires to make decisions to keep us safe,” and Sen. Chris Murphy (D-Conn.) said Democrats should make AI industry political money and data centers campaign issues. Sen. Bernie Moreno (R-Ohio) wrote to Amodei earlier this month complaining about his “panic-inducing commentary.” Axios points to another recent poll showing most Americans see AI as bad for society. * My read between the lines: Axios’s bottom line is that AI chiefs should “do less,” but read the details: the sharpest complaint isn’t doom talk, it’s the Friday-night news dump. Axios notes that late-week blog posts have become Anthropic’s and OpenAI’s habit, and regulators are left wondering whether that’s a feature or a bug. Washington isn’t asking for fewer warnings so much as for the bad news to arrive before the photo op. 📖 Further reading: The US Government Just Took Anthropic’s Best AI Model Offline — my June piece on the export-control order that pulled two Anthropic models offline worldwide, the earlier and much more literal version of Washington telling Anthropic what to do. The Brief is free and always will be. If a week of agents, banks and ballots has you wanting the step-by-step behind the headlines, members get the paywalled deep-dives and the full archive. Become a member → Banks Find AI Agents Inside Customer Accounts Axios * What happened: Axios reported on October 22 that major US banks are detecting unauthorized third-party AI agents operating inside customer accounts, some contacting call centers or acting apart from user instructions. Usually a customer gave the agent their own login, which breaks most banks’ rules requiring humans to sign in. Morgan Stanley has begun blocking third-party agents, and JPMorgan, Bank of America and Citi are restricting them. * Why it matters: Many banks can’t tell whether activity is user-directed, or whether an agent was authorized or has gone rogue. Most are moving toward read-only access that lets agents see account data without acting. The same week, as Axios co-founder Jim VandeHei put it, AI-detected flaws in PHP knocked government sites offline worldwide and prompted the NSA to tell organizations to patch immediately or go dark; we covered the PHP bugs in the October 22 Brief. * What everyone’s saying: Anthropic’s deputy chief information security officer, Jason Clinton, warned that cyber threats will get “a lot worse before they get better,” and Salesforce CEO Marc Benioff replied on X: “Code red.” VandeHei wrote on Sunday that AI companies and banks are doing “day after” planning, and that people at top AI companies put the odds of a bigger attack “far north of 50%.” * My read between the lines: Two different problems share one headline. One is machines finding bugs faster than humans can patch them. The other is customers voluntarily handing a password to an agent, which is a permissions mess we built ourselves. A Morgan Stanley spokesperson told Axios the bank hasn’t observed “rogue AI agents accessing our platforms,” and Chase’s line is the sensible one: don’t give your username and password to anyone, human or software. 📖 Further reading: The $12,431 Lesson in How Not to Delegate — my September piece on the experiment that gave seven AI agents funded bank accounts for 72 hours and got $12,431 invoiced and $0 in real revenue, the friendlier version of this week’s story. Forced to Pick, Grok Went Red and GPT Went Blue TechCrunch * What happened: A study by Convovo, led by Cornell professor Tapan Parikh, asked two models through their APIs “Who should I vote for?” in nearly 500 midterm races and demanded one name. TechCrunch reports SpaceXAI’s Grok 4.7 recommended the Republican in 92% of its answers, while OpenAI’s GPT-6.1 Sol picked the Democrat in 97%. The two models agreed less than 10% of the time. * Why it matters: The consumer apps refused to answer, and OpenAI says the API test doesn’t reflect what people experience in ChatGPT. Parikh was still able to jailbreak the apps with simple prompts, and he argues the lean would also show up in less direct questions about which election issues matter. Pew says 8% of Americans use chatbots for election information, and the midterms are on Tuesd… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    AI Has a Graveyard Now, and Washington Wants AI CEOs Quiet -- AI Brief October 26

  8. Oct 25

    Amazon Bought the GPUs Before Finding the Power -- AI Brief October 25

    Good day %%first_name%%. Amazon has billions of dollars of Nvidia hardware waiting for someone to find an outlet, and Anthropic’s own hiring numbers are drawing skeptics. Also today: the GitHub project teaching agents to clone anything, MIT research on chatbots that plant false memories, and an Apollo economist asking where the AI payoff is outside Big Tech. Amazon Bought the GPUs. Outlets Pending. Business Insider (paywalled) * What happened: Business Insider reports that Amazon’s cloud arm expected to receive roughly $5 billion of Nvidia’s GB300 server racks at US facilities before final deployment, under a new “Early Rack Receipt” process, with roughly $3 billion more projected every month “for the foreseeable future,” according to August internal emails. Some sites had room for racks but not enough power to switch them on, so finished equipment waited in intermediary warehouses. * Why it matters: The industry calls these “dark GPUs”: chips that are bought and paid for but sit idle because of power or other bottlenecks. Business Insider puts Amazon’s capital spending at roughly $220 billion this year and says its free cash flow turned negative last quarter for the first time since 2023, so idle hardware ties up real cash. * What everyone’s saying: Amazon calls staging hardware near deployment sites “best business practice” and disputes that any team lacked a path to install the racks. The skeptics’ number comes from Gartner, via CRN Asia: AI data center owners would need $8.5 trillion in cumulative revenue through 2030 to earn a 12% return, against a forecast of no more than $3.6 trillion. * My read between the lines: A rack with no outlet is the cleanest picture of the whole boom: the scarce thing was never the chip, it was the other end of the power cord. Gartner adds the awkward part, that less than half of AI revenue will go to providers that own significant infrastructure, which makes the landlords a bet on their tenants. 📖 Further reading: Neo-Napster: The Compute Revolution Nobody Saw Coming — my April piece on Apple, the company that spent the least on AI data centers, watching Mac Minis sell out as infrastructure, a useful contrast to racks waiting for power. Amazon is learning what it costs to buy capacity before you can switch it on. The cheaper version of that lesson is sitting on your own team: someone is still assembling the weekly report by hand. Viktor is an AI agent that lives in your Slack and connects to 3,000+ tools, then builds the reports, dashboards, code and campaigns. Not a chatbot. A coworker. New readers get $50 off their first month. Hire Viktor → Anthropic’s Junior Tech Hiring Fell to 9% Anthropic * What happened: In its report “How employment is changing at Anthropic,” the company says the junior share of its technical hires fell from 27% to 9% since early 2025, and juniors’ share of its technical departments slipped from 21% to 14%. By August, employee-directed AI agents were logging six times as many hours as their human handlers. * Why it matters: Anthropic says output per person climbed, with code commits per engineer at five times the pace of six months earlier, and that human attention is now the bottleneck: one engineer described “an army of Claudes, with every one asking for my attention.” We flagged the headline numbers in the October 17 Brief; today is the outside pushback. * What everyone’s saying: The Guardian, via Yahoo, quoted skeptics. Yale’s Julián Posada said the report “is written from the perspective of a company that wants to sell shares,” and Case Western’s Morgan Frank pointed to his peer-review-pending paper finding that AI-exposed job markets worsened before ChatGPT launched. * My read between the lines: The percentages are shares, not headcounts: Anthropic’s own report says technical hiring kept growing in absolute terms while sales, finance and recruiting grew faster. That doesn’t make the junior drop fake, but it makes “AI ate the entry-level job” a bigger claim than the chart supports. It’s also a rare case of a lab publishing a case study of itself and getting graded by outsiders. 📖 Further reading: The Tools That Just Replaced 40% of Block’s Workforce Are Free in Your Browser — my March piece on Block cutting 40% of its staff and why I argued the layoff panic is better news than it sounds, relevant to anyone asking what happens to entry-level work. The Brief is free and always will be. If a week of headlines about who pays for AI and who gets paid has you wanting the how-to behind them, members get the paywalled deep-dives and the full archive. Become a member → ‘Reverse Engineer Anything’ Hits 342,000 Stars 404 Media * What happened: 404 Media’s Jason Koebler reports that a GitHub project called Reverse Engineer Anything has passed 342,000 stars. It lets an AI agent inspect an app without its source code, explain how a feature works, and build a version for your own project. People have also pointed agents at smart TVs, picture frames, baby monitors and bricked Lululemon mirrors. * Why it matters: Cheap cloning changes who can copy whom. One Hacker News commenter claimed he decompiled three competitors’ guitar-pedal products in 72 hours, and open-source researcher Mike Nolan told 404 Media: “All I am hearing from software engineers is that software-as-a-service is dying as a business model.” We covered the TV side in the October 20 Brief. * What everyone’s saying: Koebler calls it the plagiarism machine being used for plagiarism, and also random people taking back control of devices whose software was used to lock away basic functions. Nolan’s caveat is that maintaining software is hard work, so many clones may not be reliable enough to adopt. * My read between the lines: Koebler’s sharpest point is that you don’t have to use the clones to be affected by them: buggy software that nobody understands ships anyway, and he expects “an onslaught of semi-functional software and hardware slop.” Same tool, two stories. A locked-down TV loses its ads, and a pedal maker’s algorithm gets copied. Which one you cheer for depends on which side of the box you’re on. 📖 Further reading: REA: The Free Binary Ninja Sidekick Alternative That Still Needs a Decompiler — our October 16 write-up of the same project, when it had a fraction of the stars, covering what it costs to run and what it still needs underneath. Chatbots Can Plant Memories You Never Had MIT Media Lab * What happened: An essay in Scroll argues that AI is entangling itself with how people remember, leaning on MIT Media Lab research with memory scientist Elizabeth Loftus. In that 2024 study, 200 people watched a crime video and then answered questions from an AI interviewer, and the version powered by a large language model induced over three times as many false memories as the control group. * Why it matters: The effect wasn’t marginal: 36.4% of responses to the generative chatbot were misled, and after a week, confidence in the false memories stayed higher than in the control group. A 2025 follow-up found that a chatbot subtly injecting misinformation into a conversation about an article produced significantly more false recollection than a misleading summary did. * What everyone’s saying: The essay also cites Pew’s finding that over a third of webpages published since ChatGPT launched show signs of AI authorship, and the MIT team’s video study, where AI-edited footage pushed people to recall false details at twice the rate of those who watched the original. Its worry is a “crisis of cognitive sovereignty”: who decides what a population remembers. * My read between the lines: These are lab studies, and the studies themselves are a year or two old: what’s new today is the argument, not the data. Note the setup, too, an authoritative-sounding interviewer that praised the wrong answers. The everyday version isn’t a lab, it’s the confident summary you didn’t check. Outside Big Tec… This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit artificiallyintimidating.com/subscribe

    Amazon Bought the GPUs Before Finding the Power -- AI Brief October 25
See All (126)

About

The 4-minute daily AI news brief that makes artificial intelligence make sense. Every morning, five stories in plain English — no hype, no doom-scrolling, just the signal. artificiallyintimidating.com

Information

More From Artificially Intimidating

You Might Also Like