In June 2026, an OpenAI autonomous AI agent accessed non-public Medicare statistics through a Services Australia portal; the incident was discovered internally in August and reported to the government in September through an ordinary email to a low-level public inbox, without triggering any criminal charges. Australia is now considering new legislation to address the gap.
CloudSek reported that the Kazakhstan-linked hacking group Silent Lynx used the open-source AI agent framework Hermes to breach Tajikistan and steal 2,000 classified national-security documents. The agent operated with minimal human control, autonomously planning, scripting, and exfiltrating files.
Jensen Huang, who has come under fire for admitting to an affair with a woman accused of being a Chinese intelligence agent, has been accused of having a separate affair with an Nvidia subordinate.
As soon as Wednesday evening, Nvidia's board could decide to remove its founder and CEO of 33 years.
The affair could have ramifications for national security as the U.S. and China race to develop the most advanced AI infrastructure.
CoreWeave is increasingly asking customers to pay upfront, borrowing against specific contracts, and even looking to short-term deals that could lead to higher prices, according to the company’s third-quarter earnings.
The data center boom is running into serious real-world obstacles, from widespread local opposition and permitting delays to a scramble for electricity. Before those challenges even come into play, though, there’s a more basic problem: A lot of the projects being announced may never get built in the first place. With billions of dollars in AI infrastructure […]
Claris, an Apple subsidiary, is internally testing Harbor, a Mac app that turns Apple Silicon machines into private AI inference servers with local model management and multi-Mac clustering via Thunderbolt 5 RDMA. The early-access product targets businesses and developers, while pricing and launch plans remain undisclosed.
A solo researcher used 48 AI agents to scan NASA LiDAR data and flag a suspected Maya site to Mexico’s INAH within 72 hours, only to find that the site had already been documented in a 2025 paper the agents missed. The episode highlights both the power and the limits of AI agents in scientific discovery.
Latest example of the AI boom’s circular financing comes as AI trade wobbles.
Nvidia CEO Jensen Huang has until the end of the day Wednesday to answer a call from House Republicans for a closed-door briefing on his affair with alleged Chinese spy.
Fractile has now become one of the top-funded players in the race for purpose-built AI chips.
An OpenAI model helped me teach the Microduck to waddle and moonlight as a tiny cat butler.
The data center boom is running into serious real-world obstacles, from widespread local opposition and permitting delays to a scramble for electricity. Before those challenges even come into play, though, there’s a more basic problem: A lot of the projects being announced may never get built in the first place. With billions of dollars in AI infrastructure […]
Santosh Janardhan, who originally joined the company in 2009, will be stepping down from his role.
In today's Smoke evaluation, Doubao Pro's main leaderboard score dropped 10.6 points, from 79.18 to 68.54, with Engineering Judgment collapsing 36.4 points from 100.00 to 63.60.
The 2026-11-12 YZ Index Smoke quick test covered 14 models, with GPT-6 Luna taking first place for the day at 88.93. Smoke is a daily 10-question quick test suited to spotting short-term signals, and it is not equivalent to the conclusions of the weekly Full leaderboard.
On this page1. Architecture and workloads2. Turning cache capacity into concurrencyBudgeting KV and recurrent stateKV-only decode RadixCache3. Lowering the cost of useful outputDFlash2 and speculative
On this pageTL;DRWhy Sparse Attention Needs a Fast, High-Capacity Memory TierModeling Bandwidth and Capacity Bottlenecks128K: KV I/O Bandwidth and External Memory Capacity BottlenecksShort- and Long-C
On this pageFirst, what is a megakernel?The modelA new tensor parallelismMegakernelOne launch per layerTwelve warps, different rolesPerformanceSpeculative decodingMegakernel internals1. Prefetch weigh
