Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Saudi Arabia's HUMAIN unveiled the Arabic-language model humain-m3, commissioned from MiniMax, on September 3, and France's Mistral AI raised €3 billion at a valuation exceeding €21 billion, developments the Center for a…
A developer detailed a "Constraint-First" prompting workflow for Cursor after a race condition in a Next.js 15 project led Claude 3.5 Sonnet to introduce three new bugs while attempting fixes. The developer's method forc…
A laid-off software developer used the 83 topics in Dan Vanderkam's book "Effective TypeScript" as hints fed to the opencode LLM agent to convert a personal JavaScript project to TypeScript, removing roughly 80% of the t…
Apple's iOS 27, shipping September 14, gives every app targeting it free access to an on-device 3-billion-parameter language model through the Foundation Models framework, with no API key or inference bill required. The …
A new npm package called 190proof provides a unified interface for calling models from OpenAI, Anthropic, Google, Groq, OpenRouter, and AWS Bedrock through a single `provider:model-id` string format. The package handles …
Anthropic alleged in a September 10th threat intelligence report that Moonshot AI secretly routed almost 300,000 Kimi customer requests to Claude through 5,380 fraudulent accounts over a 10-day period and presented the a…
Anthropic's new threat intelligence report documents eight months of Claude abuse, including Chinese AI labs Alibaba's Qwen team, DeepSeek, and Moonshot AI relaying requests en masse or extracting training data, with Qwe…
OpenAI, Anthropic and Google employees publicly confirmed they believe AI could soon kill everyone, in a series of statements collected after Jacob Coxon's warnings. OpenAI researcher Tomek Korbak said "neither anthropic…
Swiss company SOKKAN Inference launched a sovereign "Swiss" inference tier running entirely on four Intel Arc Pro B60 GPUs in Meyrin, Geneva, with no NVIDIA hardware, serving gpt-oss-20b, Qwen3-Coder-30B-A3B, and Qwen3-N…
A developer building an LLM-powered customer support agent has documented four deterministic failure-mitigation patterns that avoid asking the model to be correct, including score-threshold filtering to suppress weak sea…
OpenAI has temporarily paused new subscriptions to its $200 ChatGPT Pro plan after demand for its newly launched Astra model strained its systems, product manager Thibault (Tibo) Sottiaux announced on X. Sottiaux said th…
Context engineering — curating the system prompt, tool definitions, message history, and accumulated tool results a model sees at inference time — is a distinct discipline from prompt engineering, and failing to manage i…
A developer traced the Model Context Protocol's Sampling feature, in which an MCP server can send a `sampling/createMessage` request up to the client asking it to run an LLM completion on the server's behalf. The writeup…
A developer building a local agent harness that distributes sub-tasks to models within a 64K context window is asking the community to choose between the B70 and R9700 GPUs for the setup. The harness sequences its model …
ClickHouse released a family of built-in AI Functions that call an LLM or embedding provider directly from its SQL engine, with the full set available in version 26.8 after rolling out across releases 26.4, 26.6, and 26.…
Anthropic's September 2026 threat intelligence report, "Detecting and countering misuse of AI," disclosed that a weapons engineering cell in northern Yemen used Claude Haiku, Sonnet, and Opus models to develop missile gu…
Shopify is abandoning React Native six years after going "all-in" on the Meta-developed framework, rewriting its mobile apps in Swift and Kotlin using a new AI-powered system. Director of Engineering and Head of Mobile M…
A developer built an automated AI career coach that uses Make.com to connect Google Forms, Google Sheets, and Google Gemini, evaluating student progress reports and writing tailored feedback directly into a spreadsheet w…
Anthropic disrupted multiple attempts to use its Claude models for high-risk biological research between late 2025 and mid-2026, according to a threat intelligence report the AI safety lab published in September 2026. Th…
A technical survey traces five rewrites of transformer attention since 2017, framing every variant as either storing less or reading less. It details how multi-query and grouped-query attention shrink the KV cache from 2…