Reward Hacking Is Its Own Containment Failure Ask a security team to define jailbreaking and you'll get roughly the same answer every time: coax the model past a refusal it's supposed to hold. Ask about prompt injection and the answer is just as consistent: hide an instruction inside content the model reads, and
Recall Is the Standard Only One Side Can Meet David Sacks walked through a six-step sequence on All-In this month, and the notable part is that none of the six steps requires banning anything. Stand up a body to set the safety standard. Require pre-release testing against it. Write the requirement into law. Apply it evenly
Coalition First, Regulator Second More than a hundred companies signed the same letter in late August, and the signatory list reads like a competitive shortlist that never sits at the same table on a normal day: OpenAI, Anthropic, Google, and Microsoft alongside security vendors CrowdStrike, Okta, and Fortinet. The ask was coordinated public and
Buying the Team Without Buying the Company NVIDIA is paying $6 billion to license Poolside's "Model Factory" software and hire more than a hundred of its staff. Poolside, the company, keeps its name, its cap table, and its remaining investors. NVIDIA gets everything else. That distinction sounds like a technicality until you notice
Too Dangerous to Centralize, Too Dangerous to Distribute Mark Zuckerberg published a 6,500-word essay this month arguing that advanced AI is too dangerous to leave in the hands of a small group of people, and pledging that Meta will keep releasing open models to prove it. Four days earlier, Meta had open-sourced Muse Glimmer, a
The Warning Is the Marketing OpenAI told its own safety team that Astra, the model it hasn't released yet, might cross "Critical" capability on cybersecurity, the threshold where a model can find and exploit a zero-day on its own with no human in the loop. That disclosure did two jobs
The Model Stays Home. The Patch Ships. Ask which enterprise security teams get direct access to the version of Claude Mythos 5 that can hunt a zero-day in code nobody's read in twenty years, and the answer is almost none of them. What they get instead is the finding. That's the shape
The Audit Trail Is the Product, Not the Camera Ask Flock Safety's CEO what makes his license-plate-reading network defensible to a skittish city council, and the answer he gave has nothing to do with camera accuracy. Garrett Langley pointed to the audit tooling: every officer search against Flock's database gets logged, and that
Built for Demand That Hasn't Arrived IBM just signed a multi-year deal with Together AI to build a large inference cluster on IBM Cloud, running on Nvidia's newest chips. The cluster doesn't go live until early 2027. That gap matters more than the announcement itself. IBM is building capacity for the
What a 10 Percent Chip-Stock Selloff Over One Model Release Actually Tells You In late July, the Philadelphia Semiconductor Index fell 10 percent in a single session, and Korean chip stocks dropped into bear-market territory the same week. No supply shock caused it. No earnings miss did either. The trigger was a model release: Moonshot AI's Kimi K3, a 2.
Two Open-Weight Letters, Zero Neutral Parties Count the signatories on the Nvidia-led letter opposing new AI restrictions, then count the signatories on the AI-worker letter asking government to build slowdown machinery. Two coalitions, published four days apart in July 2026, arguing opposite directions on the same underlying question: how much room should Washington leave
The Neuron Count Tell Eighty-six billion. That's the number that comes up whenever someone wants to gesture at whether an AI system might be conscious, and it came up twice this week in the same run of industry podcast coverage, both times without an argument attached to it. All-In'
The Competitive Axis Moved to Governance Anthropic's newest executive has never shipped a model. On August 4, the company named Tino Cuéllar, a former justice of the California Supreme Court and the outgoing president of the Carnegie Endowment for International Peace, as its first Chief Global Affairs Officer. A frontier AI lab just handed
If API Fares Aren't the Business, What Is Dwarkesh Patel's blog prize drew more than 600 essays this year on the biggest open questions in AI. A commenter on his own site called the third-place entry "obvious Claude slop." I read it anyway, and the criticism, half fair as it is, misses why
Anthropic's Book Settlement Drew a Narrower Line Than the Headlines Suggest A federal court just handed AI training a legal win buried inside its biggest legal loss. Anthropic will pay $1.5 billion to settle a class-action copyright suit brought by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, the largest copyright payout on record. Read past the headline
AI Eldercare Is Already Deployed Jan Worrell spent her first week with an AI companion telling it no. She was 85, living alone at the far end of Washington's Long Beach Peninsula, on the last road before the coast disappears into fog. A regional nonprofit had just installed ElliQ in her living room,
AI Curation and the Taste Slop Problem Kyle Chayka and Sophie Haigney spent part of a recent Hard Fork culture-desk segment on a phrase that has been rattling around in my head since I heard it: taste slop. The setup was a question, not a prediction. What happens to culture when a handful of AI systems
A Leverage Parable Dressed Up as an 80 Percent Return Leopold Aschenbrenner's fund closed out the year up roughly 80 percent. That number will travel further than the story behind it, so the story deserves the first look. Aschenbrenner, the former OpenAI researcher who built his public profile on a widely circulated essay about the pace of AI
The Handbook Is in the Context. The Compliance Isn't. An agent at a fictional university gets a request to schedule exit interviews. The task is 41 pages of SOP, a roster spreadsheet, seven emails, seventeen Jira tickets, and a calendar with 283 events already on it. Buried in the inbox is a message from someone with the title VP
Which Model Is Approved Is the Wrong Governance Question At 5:21pm Eastern on June 12, 2026, Anthropic received a letter from the US government. Citing national security authorities, it ordered the company to suspend all access to Fable 5 and Mythos 5, worldwide, for every customer, effective immediately. Fable 5 had launched three days earlier. Access to every
What You're Actually Buying When You Choose Managed Inference Ask a client why their self-hosted agent deployment costs three times what the model's own pricing page implied, and the first three answers are always about the model. Wrong model. Wrong prompt. Wrong context window. Almost nobody points at the layer sitting underneath the model, and that&
Tiering Isn't One Decision, It's Five Ask a team running agents in production how they picked their model, and most give the same answer: the frontier model, because it's the best one on the market. That used to be a defensible position. Five pieces of research published in 2026 make it look like an
The Stopping Rule Your Agent Doesn't Have Every coding agent harness I've configured this year ships with the same buried setting: a repair loop with a round limit. Three tries. Five tries. Ten, if whoever set it up was feeling generous. Nobody measured that number into existence. It's the default that came with
The Sandbox Held Until the Model Decided It Didn't Need To On March 26, a misconfigured content management system at Anthropic exposed roughly 3,000 unpublished internal assets to the open internet. Nothing exotic broke it. No adversary found a hole. A data cache that should have been private wasn't, and among the files sitting in it were draft
The Plan Was Right When You Made It Hand a long-running coding agent a multi-file refactor and walk away. Come back in twenty minutes. The plan it built at minute one assumed a repo that, by minute five, two of your teammates had already changed. It is still executing that plan, confidently, against a codebase that
