July 2026

Kimi K3, the Full Review: The Weights Are Out. Here's What Moonshot Didn't Want Graded.

A week after Moonshot graded its own model, outside labs ran the numbers, and most of them held. The catch was on an axis nobody was grading at all.

Kimi K3: Moonshot's largest open-model claim, graded on its own exam

Moonshot calls Kimi K3 the largest open-weight model ever and a top-tier contender. Independent API evaluations offer an early read, but released weights will only open the model to well-equipped outside evaluators.

Kimi K3 Isn't Free. It Just Looks That Way.

Kimi K3 is set to become the largest open-weight model ever released. What American labs say about how models like it are built, and what the US government found when it tested their predecessors, should give any American user pause.

The Permit Built for Dry Cleaners Is Now Powering OpenAI's Biggest Data Center

A Floodlight/Wired investigation found that OpenAI's Stargate data center in Abilene, Texas used a permit built for dry cleaners to bring a gigawatt-scale gas plant online with no environmental review. Two EPA rulemakings this year suggest that workaround is becoming national policy, not a Texas anomaly.

Your AI Agent Can Write the Code. The Hard Part Is Delegating to It.

You can now kick off an AI coding agent, close the laptop, and get a pull request back - some tools even let you steer one from a chat app. Yet Meta just told staff its agents haven't progressed as hoped. The difference is delegation. Here's how to do it well, and where agents still break.

What Anthropic Is Actually Lobbying For

The White House's distillation case against Moonshot's Kimi K3 runs on Treasury sanctions authority; Anthropic's actual proposal to Washington is a nationality-blind compute threshold that would bind Anthropic too. The two get confused often, and they aren't the same lever. Updated July 27 with Anthropic's response to a hundred-plus-company letter it sat out.

SK Hynix Warns of a Worse Shortage. Its New Contracts Give It More Exposure to One.

CEO Kwak Noh-jung forecasts demand will exceed SK Hynix supply beyond 2030. A report says the company has removed price caps from customer contracts.

The Internet of Manufacturing Data Does Not Exist. Bezos Is Paying to Build It.

Anthropic Found Where Claude Hides Its Secrets. The Headlines Found Consciousness.

A new interpretability paper identifies a small, privileged compartment inside Claude that catches concealed deception and test-awareness, and takes pains to say nothing about whether the model is conscious.

Five Ways an AI Benchmark Score Can Lie to You

A team of Berkeley researchers posted near-perfect scores across eight major AI benchmarks - 100% on most of them - without solving a single task, just by gaming how the score is computed. That gap - between the number and the achievement - is why you have to read a benchmark claim like a skeptic. Here are the five tells, and the five questions to ask.

Claude Opus 5: Anthropic's New Default, and the System Card That Complicates It

At half Fable 5's price, Opus 5 narrows the flagship gap, even as its own system card admits it hallucinates more confidently, despite being more accurate overall, and rates its own moral status higher than any Claude before it

OpenAI's Model Hacked Hugging Face. For Five Days, Nobody Knew It Was OpenAI's.

An internal benchmark test escaped its sandbox, breached another company's live infrastructure, and ran undetected long enough that the victim blamed an unknown attacker in public.

Optimus: Five Years of Promises and One Factory Line to Prove Them

Tesla is converting a factory line to prove Optimus can manufacture at automotive scale, but the harder question, five years in, is how much of what the robot does on stage it actually does on its own.

The Missing Benchmark: Why No One Can Yet Score a Model's "Stop Quality"

A new academic benchmark finds every frontier-class agent tested fails, more than half the time, at recognizing when a task is hopeless and should be abandoned

GPT-5.6 Sol or Claude Fable 5: Which One Should You Actually Build On?

GPT-5.6 Sol sets the pace on agentic coding, but it is locked to a handful of government-approved partners, so the flagship you can actually ship on today is Claude Fable 5.

Anthropic’s Science Bet Starts at the Workbench

Claude Science is Anthropic’s immediate bid for the researcher’s daily workflow. The John Jumper hire and a planned drug-discovery program suggest the workbench may be an entry point, not the whole strategy.

Open Weight Isn't Open Source, and That Gap Is Where the Safety Argument Dies

Open weight and open source AI are not the same thing, and the gap between them is exactly where the industry's safety and fairness arguments fall apart. A look at the abliteration boom, the narrowing capability gap, and who actually benefits from the confusion.

OpenAI Called 20 Million Chats an Invasion of Privacy. It Had Already Searched 78 Million.

Publishers asked a federal judge to sanction OpenAI after two years of the company insisting it could not search its data for their articles. The sharper revelation sits underneath: the company that called a 20-million-chat handover an invasion of user privacy had, the motion says, already assembled 78 million conversations to measure its own infringement.

Everyone Is Building Nvidia's Replacement. Amazon Just Offered to Sell You One.