Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
Snowflake released Cortex Training in preview on October 30, 2026, a serverless service that runs supervised fine-tuning (SFT) and reinforcement learning (RL) on open-weight language models while Snowflake operates the G…
A 300-page survey of code intelligence, "From Code Foundation Models to Agents and Applications" (Yang et al., submitted November 2025), reports that general models "do not robustly exploit very long inputs" and that rep…
Reuters reported on November 3, 2026 that some OpenAI employees warned company leaders and board members about operational security weaknesses for years, citing five people familiar with the company's practices. OpenAI d…
DeepSeek released DeepSeek Harness (DSH), an open-source, MIT-licensed coding agent framework, on August 13 alongside DeepSeek V4 Pro, built on the Cordis composability framework in which every component — including the …
DeepSeek published seven official Creator Mode prompts in English and Chinese on a single page with copy buttons, showing how its open source DeepSeek Harness agent app (MIT licensed, over 269,000 GitHub stars, at releas…
Two open-source, MIT-licensed plugins built by an AI coding agent (Claude Code) add stereoscopic 3D to the Nintendo 3DS versions of Minecraft and Terraria, according to a practical installation guide. The Minecraft plugi…
An MIT paper reports that a frontier model the researchers call GPT-6 Astra directly edited a small Qwen reasoning model's weights without gradient-based training, raising the Qwen model's math benchmark score sixfold. T…
An MIT paper titled "Sparks of Weight-Space Programming" claims GPT-6 Astra can read another model's raw weights, identify what it was trained on, and edit a small slice of those weights without retraining — including on…
Cloudflare launched Zero Wire on November 3, 2026, a free public feed of critical-vulnerability briefings hosted in Cloudflare Radar's security section, with each card carrying a status and last-checked date and grouping…
Google's Gemini Apps Privacy Hub, last updated October 26, 2026, states that when the Keep Activity setting is on, Google uses Gemini chats to train its generative AI models and a subset of chats is reviewed by human rev…
The open-source news aggregator Horizon, built by GitHub user Thysrael, appeared on GitHub's weekly trending page for week 45 of 2026 and now has roughly 12.3k stars, 1.8k forks and 301 commits under the MIT license. Hor…
Tencent researchers Muyao Yuan, Liyan Xu, Weizhan Zhang, Muyan Jiao, Jiangnan Li, Jinchao Zhang and Jie Zhou published LoCo, a training method that teaches a reasoning model when to stop thinking, reporting completion-le…
Researchers Rikiya Takehi, Jayden Teoh, Alex Zhang and Omar Khattab reported that editing roughly 0.016% of the weights in Qwen3.5-0.8B, using GPT-6 Astra as a gradient-free editor with no training or backpropagation, ra…
A controlled study on DeepSeek-R1-Distill-Qwen-1.5B found that Sinkhorn-aligned latent tokens add no measurable accuracy over an identically trained fine-tuning control on all 1,319 GSM8K test problems, with the latent m…
Agent Autoloop, a self-hosted hub published on GitHub as KKKKhazix/agent-autoloop, starts command-line coding agents such as Claude Code and Codex from cron jobs, GitHub issues or HTTP calls and interrupts the user only …
An Elastic-sponsored Dimensional Research survey of more than 400 cybersecurity professionals at companies with over 500 employees found that 96% of security teams now use AI, with 99.8% reporting benefits such as faster…
Redis announced the next chapter of Redis Iris, a layer of the AI stack that sits between an agent and the data and tools it needs, delivering production-ready context, memory, and messaging on Day 1. Redis said agents f…
AMD published Part 7 of its ROCm profiling guide series, an AI/PyTorch edition that applies the same profiling loop used on HIP and Fortran applications to a PyTorch training workload on AMD GPUs. The guide instructs rea…
Atmanity, an agentic AI lab founded in October 2024, launched Atmee v2, a real-time full-frame video model for live avatars, alongside Causeur, a behavior model that controls personality, emotion, timing and active liste…
Axiom released an open Lean 4 formalization of the Stacks Project on November 4, 2026, which Axiom says runs to about 16 million lines and is published under Apache 2.0 in the AxiomMath/Stacks repository, built on Mathli…