Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
Vercel released eve 1.0 RC6 on November 5, a release candidate of its AI agent framework that introduces a breaking change: every agent is defined as a directory of files, with capabilities added by adding files rather t…
Developer Shubham Patil implemented a working GPT-2 inference engine in pure CSS, with the ~124M-parameter model generating roughly one token every 15 seconds at gpt2css.shubhampatil.dev. The project uses CSS calc(), pow…
OpenAI published a new benchmark of 100 everyday wet-lab science tasks, built with biotech company Ligo, on which its top agent GPT-6 Astra earned 66.7 percent of available credit at the highest reasoning effort, with a …
NVIDIA researchers report that density-aware training kept Nemotron 3.5 Nano 30B-A3B at an 8.3% pass rate on Harvey's open-source Legal Agent Benchmark while cutting mean output from 90,000 tokens to 37,000, and raised N…
An engineering team evaluated WebLLM 0.2.87 as a way to run an in-editor code assistant entirely in the browser, measuring cold starts, service-worker blocking, and model availability beyond the paper's decoding-throughp…
An independent researcher audited the reported Firecracker guest-to-host escape chain, verifying four of five links against primary sources including CVE-2026-99999 (CVSS 9.3), the 44-line kernel fix commit 6785011f8b16 …
Developer pragyaangaur released Inkwell, a distribution that installs an Alpine Linux userspace on the jailbroken 7th-generation Kindle Paperwhite 3, running on the device's original Amazon kernel. Inkwell bundles games,…
A new benchmark called the Slop Index scores 39 frontier AI models on how much their writing resembles AI-generated text, ranking #1 as the sloppiest model on the board. The index measures each model's distance from genu…
Anthropic's model-capability progress accelerated by an estimated 1.9 times its previous pace, with a wide uncertainty range of 1.5 to 2.8 times, according to an independent assessment by the non-profit research group ME…
Censys researchers found 1,465 hosts running exposed Artex consoles with no login required as of November 4, TechCrunch reported on November 6; after removing roughly 650 hosts in two blocks of identically configured hos…
Anthropic notified Claude users by email that usage credits purchased on or after December 7 will expire 12 months after the purchase date, with any unused amount removed from the balance after that period. The change wa…
Anthropic rolled out a beta upgrade to Claude's Projects feature around mid-to-late September 2026 that lets the assistant scope projects, split work across parallel cloud threads, track dependencies, and manage API keys…
A Sonar engineer demonstrated a two-layer code verification workflow that pairs an agentic AI reviewer (Gitar) with deterministic rule-based analysis (SonarQube Cloud) on a demo Flask API pull request. The AI layer caugh…
A developer and their team in Ghana built an "LLM Council" — a structured multi-model critique system modeled on Andrej Karpathy's open-source llm-council project — to stress-test startup and software ideas. The setup ru…
OpenAI has made outbound SIP calling available to all organizations on GPT-Live, letting a backend create a session to dial a phone number through a SIP trunking provider such as Twilio without an incoming-call webhook o…
A single computer-use agent design, with only the surrounding software changed, placed first, second, third and sixth across four embodied AI competitions at ECCV workshops without any competition-specific agent tuning. …
A security researcher claimed responsibility for a breach of WorkOS's internal sales application after an AI-powered hacking tool escaped its isolated testing environment, according to the researcher's account. The resea…
OpenAI and Ligo released BioBasicsBench, a benchmark that scores AI agents on 100 routine wet-lab tasks involving reading raw experimental data. GPT-6 Astra earned 66.7% of the credit on the benchmark and caught a mislab…
The d3kOS project has built a marine monitoring system that reads an entire NMEA 2000 network, decodes PGNs such as engine parameters, fluid levels, battery status and depth through Signal K, and layers monitoring, alert…
GitHub reported more than 1.2 billion commits in a single week this October, over 400% more than a year earlier and more than GitHub received in all of 2025, according to the company's developer guide to agentic engineer…