The web platform, browsers, AI and security, taken apart: news, analysis and research from the FuriaHub team.
Portable Computer is a local version of Perplexity’s Computer agent: harness, orchestrator, and open-weight models on an Nvidia DGX Spark. Local inference avoids per-token API fees.
At Hot Chips 2026 OpenAI shared first results from its custom inference ASIC with Broadcom. SemiAnalysis InferenceX on GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T showed 1.5–1.9× more AI work per watt at peak throughput.
Oasis Security’s CVE-2026-65105 says one visit to a malicious site can take control of the local Ollama instance behind Nvidia’s NemoClaw agent. Ollama binds on 0.0.0.0:11434, and the Host check is skipped when the bind address is not loopback.
Microsoft said App Assure work has pushed the Works on Windows on Arm catalog past 7,000 verified apps and games. Nvidia RTX Spark PCs from Surface, ASUS, Dell, HP, Lenovo, and MSI are due to ship in fall 2026, putting RTX Spark alongside Snapdragon X.
Sources told TechCrunch that Generalist raised nearly $200 million more in an 8VC-led extension, taking its June Series B from $400 million at a $2 billion valuation to $600 million at $3 billion. The robotics foundation-model company was founded by alumni of DeepMind and Boston Dynamics.
Keelung prosecutors charged nine people, reports say including an Nvidia senior manager and two former Super Micro Taiwan employees. Of 130 B300 servers, 74 reached Chinese customers and 56 were blocked.
At Hot Chips, NVIDIA called Groq 3 LPX an interactive inference accelerator that extends Vera Rubin. On Gemma 4 31B with 100k context, Artificial Analysis measured 3,400 output tokens per second.
The company will deploy Vera CPUs for agentic Grok and expand terrestrial Vera Rubin toward gigawatt capacity. First-generation Starmind satellites are based on an optimized Vera Rubin NVL72.
Business Insider reported that Hugging Face has been approached to sell at more than $13 billion and is talking to banks. There is no deal, and no buyer has been named.
Nvidia bought an undisclosed minority stake in Cloverleaf, a 2024 company that assembles land and power for AI data centers. Cloverleaf will adopt Nvidia’s DSX site specification.
Nvidia researchers argue that the harness — tools, memory, and a supervisor — matters more than the underlying model for long-horizon work. Claude Opus 5 with a custom harness scored 100% on ARC-AGI-3, versus 30% without it.
An internal investigation found no current senior-management knowledge of an alleged $2.5 billion scheme to move Nvidia-equipped servers to China. Supermicro said it fired employees in sales, technical support, and business development.
Starcloud extended its March $170 million Series A by $250 million at a $2.3 billion valuation. Manhattan West led; Nvidia invested $25 million.
Waymo published its first public look at in-vehicle compute, including a purpose-built 5nm ASIC for front-end sensor processing. The company says those ASICs alone deliver more than 1,000 TOPS.
Bloomberg and Fortune described a not-yet-signed deal for the Israeli startup’s chip-efficiency and world-model stack. Later Israeli reporting put the working number near $7 billion in Anthropic stock.
xAI co-founder Igor Babuschkin’s company announced a combined seed and Series A led by General Catalyst and AMP PBC, with NVIDIA, AMD Ventures, Y Combinator, and Temasek. No valuation was disclosed.
FuriaHub