Large Language Models
OpenAI's 'Astra/GPT-6' Demos Surface Ahead of Expected Sept. 3 Launch
Purported gray-test demos of OpenAI's Astra/GPT-6 show one-shot 3D generation; a Sept. 3 launch is expected, but OpenAI has not confirmed.
Latest news, developments and reporting on Large Language Models.
Purported gray-test demos of OpenAI's Astra/GPT-6 show one-shot 3D generation; a Sept. 3 launch is expected, but OpenAI has not confirmed.
Amazon's HR leader, 25 years after last coding, used AI tools to write 100,000 lines of code and build apps, as the company promotes AI training for all employees.
On August 30, ChangXin Memory filed a lawsuit against the U.S. Defense Department, ByteDance delayed its Doubao 2.2 model, and Tesla introduced a lower-priced Model 3 in Hong Kong and Macau.
Linux Foundation IT director says AI scrapers make up most git.kernel.org traffic, forcing feature cuts and new access limits.
A new benchmark of voice-agent inference APIs shows time to first token (TTFT) is not enough; time-to-first-sentence matters more, with Baseten's gpt-oss-120b hitting 0.23s TTFT.
A study of 6,214 corporate domains found 120 unclaimed references in AI documentation that could be hijacked to deliver malware through AI agents like Claude, Codex, and Hermes.
AI-generated deepfakes of late-night TV hosts like Jimmy Kimmel are spreading on social media, fooling viewers and raising concerns about disinformation, free speech and platform accountability.
An NPR and NewsGuard experiment found AI chatbots reject foreign disinformation about 75% of the time, outperforming search engines and AI-generated search summaries.
Anthropic opens a research preview of a spec for AI agents to safely control devices, cutting setup from weeks to hours.
Experts debate neuromorphic computing at WRC 2026; Anthropic launches MHS; MirroS open-sources Code-as-World.
A global survey of 1,800 professionals finds 91% say their organization still falls short on AI, with gaps in tools and strategy. Thomson Reuters COO Kirsty Roth advises focusing on grounded explorations and specific production use cases.
Unmonitored AI agents can destroy systems and rack up huge costs, as even a cost-management vendor learned.
Senator Brian Schatz is rising as the No. 2 Democrat in the Senate, a court ruled the Pentagon's blacklisting of Anthropic illegal, and the Strait of Hormuz has been demined.
Google Flow gets Gemini Omni 1.1 Flash, enabling precise AI video editing, faster drafts, higher resolutions, and smarter scene extension.
AI agents are lifting CPU demand and agent revenue, as Nvidia shows CPU racks and Unisound reports growth.
OpenAI published Jalapeño inference chip results, Anthropic extended Claude Code's quota boost, and Cursor rolled out Origin for agent-scale code hosting, reflecting a push toward inference efficiency and long-running AI workloads.
ACE Robotics and the University of Hong Kong unveiled StreamPI, a method that gives vision-language-action models continuous temporal modeling, boosting real-robot task success by up to 33 points.
Researchers recover hidden AI reasoning from major APIs via compatible models; Anthropic adds cross-session messaging to Claude Code.
Cursor open-sourced MoK to cut MoE communication latency on Nvidia GPUs, while Meta released Muse Glimmer, a 30B local agent model with 128K context. Both tackle software-level bottlenecks in AI execution.
A Use.AI survey of 15,428 adults finds 64% of people feel more confident challenging professional advice after checking with AI, highlighting an 'expertise recession.'
Tencent Hunyuan launches open-source Hy4 preview, a 770B-parameter LLM with 1M context, ranking among top open models in real-world productivity tasks.
WRC 2026 embodied AI shifts from demos to ROI; domestic chips power commercial video generation and GLM-5.3-Flash.
Google announced two AI updates: a new speech-to-text model, Gemini 3.5 Transcribe, and a feature that lets users import purchased Play Books into Gemini Notebook.
Pew Research finds ChatGPT is shaping human writing, with doubled em dashes and a surge in words like 'delve.'