Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A developer working on a locked-down enterprise Windows machine built a Python bridge using pywin32 to let Claude Code read and draft Outlook emails without MCP servers or Azure app registration. The single-file tool exp…
The official Anthropic Ruby SDK now supports building AI agents in Rails with tool design, MCP servers, human approval gating, testing, and cost control. The SDK includes streaming, connection pooling, and a tool runner …
Claire tests GLM-5.2, an open-weight model from Z.ai, in her ChatPRD codebase, finding it production-grade and capable of replacing Opus for some coding workflows. The model benchmarks near Claude Opus 4.8 and above GPT-…
FamNest engineer discovered that AI agents fail to call APIs correctly when documentation is written for humans, not machines. The coach agent in FamNest's agent graph repeatedly constructed malformed calls to a retrieva…
The New York Times escalated its copyright lawsuit against OpenAI and Microsoft, accusing Microsoft of building a supercomputer specifically to help OpenAI steal copyrighted articles. The Times seeks to amend its complai…
AI is transforming software development by automating repetitive tasks like code generation, debugging, and documentation, but it is not replacing developers. Developers remain essential for understanding business goals,…
OpenAI is investigating Codex quota depletion after Pro plan users exhausted weekly limits in one to two days, revealing that AI coding agent adoption is outpacing capacity forecasts. Engineering lead Thibault Sottiaux c…
As recursive self-improvement accelerates, humans risk losing the ability to interpret and guide autonomous research agents, which could lead to safety failures. A new research agenda proposes studying how to keep humans…
A second-year computer science student at the University of Debrecen built RepoLens, an AI system that scans GitHub repositories for developer onboarding, screens technical candidates by analyzing commit history, and ope…
Nearly 400 local newspaper publishers filed a class-action lawsuit against OpenAI and Microsoft on June 24, accusing the tech giants of using copyrighted journalism without permission to train AI products like ChatGPT an…
Anthropic reported elevated errors on Opus 4.5, a model in its Claude product line, and is offering email and SMS notifications for incident updates.
Target developed a generative AI-based system that uses embeddings and large language models to semantically match new marketing campaigns with similar historical ones, improving forecasting accuracy and reducing manual …
The RSPCA warns that 10% of UK pet owners are using AI chatbots like ChatGPT and Gemini for pet care advice instead of consulting vets, driven by rising veterinary costs. The charity says AI cannot physically examine ani…
Apple integrated Alphabet's Gemini large language model into Siri at WWDC 2026, driving demand for higher-memory devices as DRAM prices surged over 60% from the prior quarter, according to Micron Technology. The company …
An open research program called Project UCTF has been launched to investigate whether a machine-native intermediate representation can serve as an alternative training representation for multilingual AI models, reducing …
DeepSeek will launch its V4 series in mid-July with peak hour pricing that doubles token rates during designated periods, a first for the AI industry. The tiered pricing affects both V4-Pro and V4-Flash models, with peak…
Apple unveiled a rebuilt Siri AI at WWDC 2026, powered by Apple Foundation Models and a partnership with Google Gemini, emphasizing privacy through on-device processing and Private Cloud Compute. The move positions Apple…
Two weeks after the US restricted exports of Anthropic's Mythos and Fable 5 models, Chinese lab Zhipu AI released GLM-5.2 and Japanese startup Sakana AI released Fugu, both matching or surpassing Anthropic's performance …
A comparison of AI coding assistants Cursor, Claude Code, GitHub Copilot, and Windsurf across six real engineering scenarios evaluates their performance in 2026.
A new analysis argues that retrieval-augmented generation (RAG) pipelines frequently fail in production due to retrieval irrelevance and context poisoning, with enterprise implementations showing a 72% first-year failure…