Latest
Windows 11's Advanced Shader Delivery is now self-service for developers. We explain how SODB and PSDB differ, the two ways to collect shaders, why waits can return after updates, and what the October GPU expansion and per-game delivery conditions mean.
After the US government's AI-to-SI renaming, new .si registrations jumped. Official monthly data, resale-market observations and hosting providers' responses show how acquiring a name differs from AI services actually taking off.
OpenAI dismissed three employees for violating rules on handling confidential information. We compare its public policies on external safety assessments, employee concerns and unauthorized disclosure, and look at what METR's independent investigation reveals about access and its limits.
NVIDIA has announced a 64GB DGX Spark starting at $4,999. We compare that price with the earlier 128GB model and look at KV cache headroom, the conditions for linking two units, and what NVIDIA Sync offers now and what is planned.
Micron's four-site expansion in Taiwan, broken down by wafer fabrication, HBM packaging and testing. We examine what the NT$1.6 trillion cumulative investment means, the reuse of existing facilities, and Nanya Technology's approval for a Pingtung plant, using official materials and local reporting.
Nikkei reports Toshiba will spend about ¥60 billion to double its HDD production capacity. We estimate the drives needed based on the M12's rated capacities and look at SMR adoption conditions, product development stages, and shifts in Toshiba's finances.
In Tavus's own one-minute call experiment, 26 of 54 people judged Griffin-Lite to be human. We read that result alongside NVIDIA's benchmark breakdown to separate naturalness, misidentification rates, and response timing.
Microsoft has released a streaming transcription model and two speech synthesis models in public preview. We check the primary sources on how interim text is corrected, differences from the standard version in features and pricing, how accuracy and latency were measured, and Japanese support.
Researchers disclosed Branch Target Reuse (BTR), an attack that exploits stale branch predictions left in the CPU after JIT code is reused. This article examines the Linux root-hash experiment, what was and wasn't demonstrated on Firefox and GraalVM, Intel's assessment, and the state of fixes.
Cloudflare's decision model Clef returns probabilities for predefined options so programs can branch automatically. We look at the accuracy gap between Clef and the lighter Clef-flash in published evaluations, plus what is open and when fine-tuning is available.
