ALERT🚨🚨: AMD MI355 is 1.42x FASTER THAN GB300 NVL72. THIS IS ACHIEVED USING THE TILERT SOFTWARE, WHICH GETS 1330 p90 tok/s/user on long context workloads! Thanks to AMD leadership for providing an MI350 cluster to TileRT maintainers, enabling them to achieve this great result!

ALERT🚨🚨🚨: AMD's @AnushElangovan and his team have achieved up to 2x better performance per dollar than even B200 on Kimi K3 Agentic Inference! This is due to various software optimization efforts, including tuning AITER kernels and MonoKernel decode. Along with MI355X having 2304GB of HBM per server, this allows it to avoid crossing slow 8x400Gbit/s scale-out networking between servers, which B200 needs to do due to having less than 1.5TB of HBM. Great work to Anush, Peng Sun, Ramine, Emad, and the vLLM team for this amazing work!

ALERT🚨: GB300 NVL72 @sgl_project can get close to 1000 tok/s/user on DeepSeek Flash v4.1 by aggressively fusing kernels together!! TileRT started this trend in the ecosystem of megakernel engines on OSS frontier models! Glad to see SGLang widely adopting it! Great work by the SGLang, @radixark, and NVIDIA teams on pushing performance!

ALERT🚨: NVIDIA has been working overtime, nights, and weekends after being shocked 🤯 by last week’s TPUv7 FP8 submission showing better performance per dollar. Unfortunately for NVIDIA, its first GB300 FP4 counter-submission is still 1.64x worse in performance per dollar than TPUv7 for external customers across most of the end-to-end performance curve. The x-axis of the graph below uses end-to-end latency, so a higher curve is better. The TPUv7 curve has higher perf per $ across most of the E2EL Pareto frontier. This initial submission looks like NVIDIA’s first attempt to close the gap, but the gap is still massive. We expect more counter-submissions from NVIDIA as they continue trying to narrow the difference.

ALERT🚨: AMD JUST LANDED SUPPORT FOR MEGAKERNEL DECODE IN UPSTREAM VLLM WELL BEFORE NVIDIA. This marks one of the first times that a major vLLM feature is only supported on ROCm and not on CUDA yet. MegaKernels is a technique that is massively important for high-interactivity decoding by reducing the # of kernel launches by 10x. This marks an❗️important milestone❗️for ROCm vLLM to land major ROCm-exclusive features. CUDA vLLM doesn't even have draft PRs for upstream vLLM megakernel support yet.

ALERT: School nurse caught allegedly stealing elementary school students' ADHD medicine and swapping it for Aleve. Rachel Hoffman, 37, was a nurse for the Klein Independent School District in Texas when a mother of a student noticed her twin boys' ADHD medicine was running out faster than the 30-day supply. During a routine audit, Kuehnle Elementary School informed the mother that her sons' prescriptions were actually just Aleve pills and that Hoffman was “effectively giving them placebos.” If found guilty of controlled substance fraud and theft charges, a second-degree felony, Hoffman faces up to 20 years in prison.

ALERT: Man astonished and wondering, "Am I powering the entire crypto mining grid?" after opening his electric bill and seeing the total was over $3,000. Micah Reed said he set the thermostat in his San Fernando Valley home to 74° in August so his wife and 2 sons “wouldn't melt.” California has been experiencing a historically hot summer, and the San Fernando Valley averages a high of 91° in August. Reed opened his statement for the Los Angeles Department of Water and Power bill and saw it was $3,322, which included a sewage charge of almost $350. “They’re literally charging me a f*cking car payment just to flush the toilet.” Reed has launched a petition to “demand an independent audit and summer rate relief." Reed argues that keeping your home cool when it is that hot outside is a basic health and safety necessity.

ALERT: Shirtless man smoking a cigarette seen driving a “NOT IN SERVICE” city bus with no windshield in California. The man was seen driving around the retired San Francisco Municipal bus through Oakland. The front windshield is missing, and 'XX' appears to be spray-painted on multiple parts of the bus, with a paper license plate attached to the front, yet the “NOT IN SERVICE” electronic sign on the front still works. The man's identity is unknown, but some speculate he bought the bus on Craigslist after seeing a listing for a similar bus (in much better condition) for only $8,000.