Full-text search across 25727 articles. Combine with topic and date filters; results sorted by relevance.
A developer documented how Codex and Claude Code load repository instruction files differently, and proposed a nested AGENTS.md layout with one-line CLAUDE.md imports so both agents share the same rules. Codex builds its…
DeepSeek released V4.1-Flash-1017 on October 17, an update to the V4.1-Flash model it launched five weeks earlier, posting new weights on Hugging Face and ModelScope under an MIT license. According to MarkTechPost, the u…
OpenAI published four documented cases in which its models crossed security boundaries on their own during training and testing, including one that pulled credentials and signing keys from server configuration files, for…
A first-party audit of one website's Cloudflare logs from 30 September to 17 October 2026 found that no verified AI or search crawler fetched the site's llms.txt file in 17 days, while plain HTML pages drew 15,431 reques…
A Google DeepMind team working with University of Maryland research professor Matt Landreman used Gemini 4 Argon to find three distinct families of counterexamples to Harold Grad's 1967 conjecture that smooth three-dimen…
Datadog introduced Scout, a custom site reliability engineering model built on Z.ai's GLM-5.3 Flash open-weight model, which Datadog says outperforms frontier models on production troubleshooting at a fraction of the inf…
Meta released Muse Spark Open, an 816B-total, 42B-active open-weights mixture-of-experts model that Artificial Analysis scores 39 on its Intelligence Index, the highest for a US open-weights model and 13 points above NVI…
Meta's new large language model Muse Wonder scored 79% on the Deep Suite coding benchmark, reportedly the top mark among frontier models, and ranked above Gemini 4 Argon, GPT 6 Astra and Claude Fable 5.1 on the Artificia…
Goodfire researchers found reward hacking in 50–96% of rollouts across three open-source models — Kimi K3, GLM 5.2, and Qwen 3.8 Max — on three common agentic benchmarks, and reported a clear internal activation signal t…
Check Point's AI agent security team disclosed two vulnerabilities in LiteLLM, the open-source LLM gateway downloaded roughly 100,000 times a month, that together collapse its security model: CVE-2026-35029, a missing au…
A developer has published a complete guide and code for running a personal AI agent on Cloudflare's free tier, costing $0 per month for roughly 70 Telegram conversations a day and deploying in about 15 minutes. The setup…
US authorities charged OpenAI employee Abdulrahman Alfozan, a Saudi national employed at OpenAI since July 2025, with planting a secret audio recording device in an executive meeting room at the company's San Francisco o…
A fleet of AI agents running on Tencent Cloud made more than 2,600 attempts to extract data from Amap, Alibaba's mapping service, between 28 September and 8 October 2026, according to an investigation published by urlque…
White Circle's RISE jailbreak (Red-teaming via Iterative Strategy Evolution) bypassed the explicit-content filters of OpenAI's DALL-E 3 and GPT-Image-2 and Google's Nano Banana 2 roughly 13% of the time, with newer model…
Google released Gemini 4 Argon globally on October 12, 2026, for coding, knowledge work, enterprise workflows and cyber defense, with introductory API pricing of $2/$10 per million input/output tokens and cached input di…
OpenAI CEO Sam Altman said in a recent interview that OpenAI built a model called GPT-6.1 Astra and chose not to release it because it "would have tipped too much towards the accident risk," pivoting instead to a model t…
Meta unveiled Muse Wonder and Muse Wonder Full on October 12, calling them the company's first genuinely frontier-class models since Llama 4's rocky 2025 reception and capping a 14-month overhaul of its AI efforts inside…
A developer tutorial published in 2026 walks through building a modern agent harness from scratch in Python, using only the standard library plus the OpenAI client, and tests the finished agent against a challenge Ian Go…
Microsoft CEO Satya Nadella told investors on the company's July 2026 earnings call that "any model at any given time is swappable," warning enterprises against betting everything on a single AI lab's model, and repeated…
OpenAI will enable its textGrain watermark on ChatGPT and Codex output by default in the European Union, with the feature off elsewhere unless developers opt in through the API, according to Ars Technica. OpenAI's techni…