Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
The U.S. Commerce Department has lifted export controls on Anthropic's Fable 5 and Mythos 5 AI models after the company agreed to implement new guardrails and classifiers to prevent jailbreaks. The restrictions were impo…
A developer testing self-hosted AI agents for credential leaks found that while the final answer leaked a planted API key only 0.5% of the time, the model's hidden reasoning leaked the same key about 26% of the time, and…
Habitual publisher traffic is collapsing as AI bot traffic surged 187% in 2025 while human traffic grew only 3.1%, eroding the click-based value exchange. Direct traffic to major publishers fell 19.9% to 33.1% over three…
Anthropic faces allegations of embedding spyware-like code in its Claude AI model to detect connections linked to China, raising compliance and safety concerns that may delay the restoration of Claude Fable 5 in the US. …
The Trump administration restricted access to Anthropic's Mythos 5 and Claude Fable 5 AI models for foreign nationals on June 12, citing national security concerns. Anthropic disabled both models entirely to comply, lead…
Anthropic launched Claude Sonnet 5 on June 30, bringing agentic capabilities like browser use, terminal work, and multi-step tool execution to its cheaper Sonnet tier and making it the default for Free and Pro users. The…
A developer released an open-source tool called Voice-to-SQL that uses an LLM (Llama 3.3 70B via Groq) to convert plain English business questions into SQL queries against a sample SaaS database. The tool shows the gener…
A growing gap between insider and outsider access to frontier AI models threatens the effectiveness of external safety research, with government restrictions and deployment lags likely to widen the divide. The author arg…
In 2026, the open-source automation tool n8n, with over 80,000 GitHub stars, enables users to create intelligent workflows integrating AI without advanced programming. A developer demonstrates building a workflow that re…
AI agents in data engineering should be structured with a deterministic correctness layer to ensure reproducibility and trust, according to a new analysis. The article outlines three levels of AI agents—chat-phase, auton…
Z.ai launched GLM-5.2, the first open-weight model to top the HuggingFace Open LLM Leaderboard and beat proprietary coding products on SWE-bench Verified with a 72.3% score. Trained on Huawei Ascend hardware under Apache…
Anthropic launched Claude Sonnet 5 on July 1, 2026, with autonomous agent capabilities, Computer Use 2.0, and 40% lower pricing, while the US lifted export controls on its frontier models Fable 5 and Mythos 5. The dual a…
Apple confirmed a $1 billion/year deal to use Google Gemini as the foundation of Siri across 2 billion devices, replacing OpenAI's ChatGPT. The partnership ended due to OpenAI's latency issues, poor subscriber economics,…
Marco Bambini proposes a new database architecture designed to serve both human users and AI agents, emphasizing natural language interfaces and semantic understanding over traditional query languages. The approach aims …
Researchers introduced Pollux, a new class of decoder-only LLMs that use native Leech-lattice quantization to achieve 0.76 bits per parameter, compressing a 1B-class model into 76 MB of SRAM. The model abandons continuou…
DeepReinforce released Ornith-1.0, an open-source model family that outperforms Claude Opus 4.7 on Terminal-Bench 2.1, challenging frontier AI labs.
AI-powered search is raising the bar for business websites, requiring companies to move beyond mere online presence to structured, fast, and useful sites that help both users and AI systems understand their value. Silico…
Meta is developing a cloud infrastructure business called Meta Compute to rent out unused AI computing power, potentially competing with AWS, CoreWeave, and Lambda. The initiative, still in early stages, could turn idle …
Intel's new ASUS NUC 16 Pro mini PCs, tested in a cluster of three, failed to speed up large language model inference, with token generation dropping to half the single-machine rate due to memory bandwidth limitations. T…
Apple CEO Tim Cook held a virtual meeting with EU tech chief Henna Virkkunen on June 30 to discuss the delayed launch of Siri AI in the European Union, which Apple attributes to regulatory hurdles under the Digital Marke…