Panduit IntraVUE now faces a CISA cybersecurity warning covering five vulnerabilities, including one with the maximum CVSS 10 severity score. The most serious flaw can turn the monitoring software into an active proxy that bypasses operational technology network segmentation. That finding changes the risk calculation for industrial operators. IntraVUE is intended to improve visibility across industrial Ethernet networks. The vulnerable application can instead provide a route

AMD has detailed a 256-core EPYC 9996 processor that challenges Nvidia Vera and Intel Xeon with unusually aggressive performance claims. The amd nvidia contest now extends beyond accelerators into the CPUs that organize AI workloads, feed GPUs, and run agent tools. The Zen 6 processor, code-named Venice, combines 256 Zen 6c cores, 512 threads, and 1,024 MB of L3 cache. AMD says it delivers up to 3.4 times the performance of Intel's 128-core Xeon 6980P in a selected AI benchma

AMD has extended its server CPU roadmap to 2030, despite the industry still waiting to test its newly launched Zen 6 generation at scale. The amd tom roadmap coverage centers on Zen 7 "Florence," scheduled for 2028, and Zen 8 "Ravenna," now under development for 2030. AMD presented both at its Advancing AI event on July 23, 2026. Florence will arrive with new AI compute extensions, a next-generation manufacturing process, and support for newer memory technologies. It will als

China’s Ministry of Transport has reportedly joined another major policy push, despite unresolved questions about standards, oversight, and implementation. The rsshub 36kr alert says the ministry and the National Development and Reform Commission issued a reform opinion covering China’s integrated transportation system. According to the July 24 report, the opinion calls for faster development of “AI plus transportation.” It also addresses digital infrastructure standards, int

Xiaohongshu HELMSMAN has moved production vector search onto about 40 all-flash servers, replacing workloads that consumed roughly 35,000 CPU cores and 350 TB of DRAM. The company says the change reduced associated hardware costs by more than 90%. That result challenges a familiar infrastructure assumption. Low-latency vector retrieval has usually depended on graph indexes held almost entirely in expensive DRAM. Flash-based systems reduced memory requirements, but their slowe

Xiaohongshu open-sources BigMac with a striking claim: its new training pipeline runs multimodal workloads 1.08 to 1.9 times faster than tested baselines without letting activation memory grow with batch size. The system targets a stubborn conflict in multimodal model training. Teams usually gain speed by retaining more intermediate data, or conserve memory by accepting more idle GPU time and communication. BigMac does not introduce a new foundation model. It changes how an e

Anthropic’s $1.5 billion copyright settlement received final approval after authors accused the company of obtaining pirated books used while developing Claude. Yet the outcome does not declare AI training itself unlawful. That distinction makes the agreement far more consequential than its headline number. U.S. District Judge Araceli Martínez-Olguín approved the class settlement on July 20, 2026. The agreement covers 482,460 eligible works, with approximately 91 percent clai

Tencent Hunyuan released the Hy3 model. The system uses a 295 billion parameter mixture of experts design. Only 21 billion parameters activate during inference. This efficiency matches much larger models. Hy3 aims at agent workflows rather than general chat. It already runs inside WeChat for over one billion users. Release Details Meet Real Workloads The company tested Hy3 on more than fifty internal business tasks. Task success rose from 72 percent to 90 percent. Average tas

DeepSeek-V4 Flash completed its first verified reinforcement learning runs on AMD Instinct MI355X GPUs using the Miles framework. The 284 billion parameter mixture-of-experts model activates 13 billion parameters per token and finished more than 100 optimizer steps across four eight-GPU nodes. The test combined SGLang for rollout generation and Megatron for policy updates while Miles handled asynchronous loops and weight synchronization. Training and rollout log-probability d

Anthropic and AE Studio introduced a new technique called GRAM that adds removable modules inside large language models. The modules hold specific dual-use knowledge during training. Teams can delete the modules later to strip out those capabilities. GRAM stands for Gradient Routing Auxiliary Module. Engineers insert a small set of extra neurons at every transformer layer. During training the model learns to route topics such as virology, cybersecurity, nuclear physics, and s

Robbyant has released LingBot-VLA 2.0, an open source 6B parameter vision language action model built to control multiple robot platforms through one unified representation. The model uses a MoE action expert design on top of a Qwen3-VL-4B-Instruct backbone. It encodes robot states and actions as 55 dimensional normalized vectors. This approach allows the same weights to drive different robot bodies without separate fine tuning for each hardware type. Training data reached ab

A father created a simple communication tool for his non-verbal autistic son. He tested it in a speech therapy waiting room three weeks ago. Mothers of other non-verbal children started crying. The therapist cried for five minutes straight. That moment revealed clear product-market fit. The app targets children who struggle to understand language. It differs from traditional AAC devices built mainly for adults with physical limits but intact language skills. The demo that cha

Recent outages on Xbox Cloud Gaming and GeForce Now highlight the fragility of subscription-based cloud entertainment when infrastructure fails during peak hours.

Xiaomi set a new benchmark with its autonomous lap at the Nurburgring. The YU7 GT completed a full unmanned timed run in 10 minutes 29.483 seconds. This marked the first official autonomous entry on the track's leaderboard. The result came from months of track-specific data rather than added layers of agent logic. The lap exposed a key pattern in AI performance. Reliable output in difficult conditions depends on training that covers unusual situations. Xiaomi collected high-f

Humanize PPT v0.9 released an open source workflow that puts audience context ahead of slide visuals. The update processes outlines through abstract syntax tree logic, adds real preview pages, and folds speaker notes plus QA fixes into one flow. See the Humanize PPT v0.9 release notes and the AST logic documentation. Independent coverage in The Verge and a Bloomberg analysis noted the shift toward reusable meeting context over visual templates. Most AI deck tools still start

DeepSeek raised more than $7.4 billion in its first external round at a valuation above $50 billion. The company kept founder Liang Wenfeng in control through a limited-partnership structure. Investors accepted the terms because they see model cost compression as inevitable. The durable advantage now sits in the quality of work context that agents can reuse across meetings, documents, and prior decisions. Control structure changes investor expectations Liang Wenfeng retained

Physical AI robots attracted major funding in early 2026, but real-world deployments continue to expose large gaps between controlled lab demonstrations and unpredictable factory or home conditions.

Enterprise buyers now demand measurable proof of AI governance controls during procurement, shifting RFPs toward operational evidence and live demonstrations rather than policy promises.

Neuralink reported new human trial results for its brain implant this month. The device now supports longer sessions of direct neural control over external software. Patients completed tasks that previously required multiple manual steps. The move accelerates competition in bio-integrated AI. Several firms are racing to match signal stability and reduce surgical risk. Regulators in the United States and Europe have begun reviewing expanded safety data. Primary opponent pressu

Is the word "expect" making your forecasts sound weak or your requests sound demanding? Discover powerful synonyms for leaders, consultants, and analysts to anticipate, require, and forecast with greater strategic precision.