A new architecture called BatchDAG reduces LLM API calls by up to 47x through entity-aware batching, enabling queries over 50,000+ meetings in under 60 seconds at $0.02–$0.24 each, according to a prep…
Google's latest Gemini models deprecate and ignore the temperature, top_p, and top_k API parameters, removing manual control over output randomness. Developers relying on these parameters for determin…
Alibaba has released Qwen-Image-3.0, a model capable of natively rendering precise multi-line text, structured charts, and accurate document layouts directly into generated images, removing the need f…
Google launched Gemini 3.6 Flash, priced at $1.50/$7.50 per million input/output tokens, which reduces output token consumption by 17% and requires fewer reasoning steps and tool calls per task compar…
A federal judge has approved a $1.5 billion settlement against Anthropic for using pirated books to train its Claude models, establishing a costly precedent for training-data provenance. The penalty, …
OpenAI has launched ChatGPT Work, a packaged offering for small businesses that provides native workflow automation and skill-building tools, signaling a push to standardize AI adoption below the ente…
A tool-using fact-checking agent achieved 97.8% selective accuracy on answered claims by adding an "uncertain" abstention verdict, deferring just 6 of 95 cases, while overall accuracy remained at 91.6…
The open-source agrepl framework, a Go CLI tool, achieves 98.3% median latency reduction for AI agent replay by intercepting all external transport-layer interactions via a local MITM proxy, guarantee…
A new arXiv paper reveals that PlanFlip attacks achieve a 0.68 attack success rate on GPT-5, with homogeneous multi-agent pipelines using the same LLM backbone for planning and auditing being highly v…
OpenAI has released lessons from deploying long-running AI models, warning that autonomous agents introduce failure modes such as goal drift, compounding errors, and unsafe intermediate actions that e…
Researchers from Yang AI Lab released Inertia-1, a motion foundation model pretrained on over 18 million hours of unlabeled accelerometry data that generalizes zero-shot to unseen body placements, sam…
Chinese open-weights models from Moonshot and Alibaba now match OpenAI/Anthropic frontier quality at a fraction of the inference cost, with roughly 80% of startups already using a Chinese model in the…
AnovaX, a new voice assistant described in a preprint on arXiv, runs entirely locally on a user's computer using a single Python process and Gemini-generated JSON plans, avoiding cloud-orchestration f…
Simon Willison built an app that highlights ten common clichés characteristic of LLM-generated writing, enabling teams to more easily detect and mitigate over-reliance on predictable phrasing in AI-as…
NVIDIA NeMo Automodel now integrates with Hugging Face Diffusers, enabling scalable fine-tuning of video and image models like FLUX.1-dev and HunyuanVideo without checkpoint conversion or model rewrit…
Apple has filed a lawsuit against OpenAI, alleging that over 400 former Apple employees now work at the AI company and that OpenAI engaged in trade secret theft, according to a TechCrunch report. The …
Moonshot AI's Kimi K3 open-source model has become the first to match frontier proprietary models like Claude 5 and GPT 5.6 on independent benchmarks, achieving frontier-level performance and slashing…
Meta is in talks to lease computing power worth up to $10 billion to Anthropic, a deal that would allow Anthropic to scale its AI model training without massive upfront capital expenditure. The arrang…
Capital One open-sourced VulnHunter, an AI-powered code security tool that applies attacker-perspective analysis to source code, identifying exploitable defects and proposing targeted remediations. Th…
OpenAI CFO introduces a scorecard for measuring AI ROI, featuring metrics such as cost per successful task and return on compute, enabling engineering teams to quantify the value of LLMs and agents in…