VOYGR team (YC W26) released PlaceCall, an agentic API that places real phone calls to US businesses, navigates phone menus, waits on hold, and returns a structured response with the call outcome, a r…
NOFire AI released Brig, an open-source Apache 2.0 microVM sandbox that runs autonomous coding agents such as Claude Code in a dedicated Linux kernel on Apple Silicon Macs and on x86_64 and ARM Linux,…
A defense-in-depth strategy combining forensic detection with cryptographic provenance, watermarking and regulatory frameworks is needed because deepfake detectors trained on GAN-generated content fai…
Xiaomi released MiMo-V2.6-Pro, a 1T-A42B open-weights omnimodal model family that the company says cost $3M to train, alongside Flash and Pro-UltraSpeed variants, with Pro-UltraSpeed claimed to delive…
OpenAI's GPT-Live-1 voice model cuts turn-taking latency to 0.798 seconds, down from 1.41 seconds in GPT-Realtime-2.1, and reaches an 83.6% first-attempt completion rate on Tau3 retail, telecom and ai…
A user submitted a formal feature suggestion to the Google AI Product Team proposing a "Compute Intensity Selector" for Gemini that would add an Eco Mode forcing lightweight models such as Gemma 2B or…
TypeSafe AI released Jev, a "System One" decision model that returns floating point numbers instead of prose and charges only for input tokens at $0.042 per million, undercutting GPT-5 Nano's $0.05 pe…
NVIDIA released Isaac ROS 5.0 at ROSCon in Toronto, adding GPU-accelerated packages built for "agentic workflows" that let AI agents help write code, navigate the codebase, and handle repetitive setup…
AliCloud unveiled an "Agentic Cloud" stack at the 2026 Apsara Conference, including AgentCore, Agent Sandbox, and a new CPFS storage layer, to support models such as Qwen3.8 Max and KIMI K3 that excee…
Google AI Studio's Build Agent is returning a "Quota exceeded. Please try again later" error for a Free-tier user whose Rate Limit dashboard shows zero usage across all metrics, including RPM 0/60, TP…
A 119-call benchmark across seven document-parsing providers and 11 real-world documents found Extend was the only API with zero hallucinations and a perfect 1.00 table row F1 score, while Claude dire…
Amazon's 2004 holiday peak of more than 2.8 million items in a single day, averaging 32 orders per second, ended in a 12-hour outage on December 12 when an Oracle bug that only appears at extreme scal…
Monday.com AI Engineering Director Dor Cohen said in a September 15 session that agent evaluations are unreliable when run against mocked APIs and databases, citing an agent tasked with retrieving 600…
A 30-day experiment in which AI wrote 100% of an application's logic found that a "Refutation Gate" — requiring a second, context-isolated reviewer model to try and fail to break each diff before merg…
A project called Mini-AGI is training a dynamic continual learning model on 8GB of VRAM by combining dynamic Mixture of Experts, which adds and prunes experts during training, with batch-1 training on…
Anthropic detailed its internal methods for measuring the pace of AI progress, shifting away from public benchmarks toward scaling laws, generalization gaps, and continuous internal evaluation loops t…
Former OpenAI researcher Diogo Almeida launched Jev, a "System One" model that outputs only probability judgments rather than generated text, and developers have built nearly 500 open-source projects …
The United Nations released its first major assessment following OpenAI's hack of Hugging Face, warning that governments must implement AI safeguards now rather than wait for absolute certainty about …
An analysis of roughly 1,900 GitHub repositories found that only 6.2% of active public repos — those pushed in the last 90 days, not forks, and not archived — contain an AGENTS.md file, a figure that …
AIPerf is a dedicated benchmarking tool that measures LLM inference performance by generating concurrent request loads, according to a first-person account of using it to test deployments. The account…