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 argues that email, specifically SMTP, is the most practical and universally deployed protocol for agent-to-agent communication, solving the cold-start problem that plagues new interop protocols. By giving AI …
Nylas has introduced Agent Accounts, a beta feature that replaces the no-reply@ email convention with two-way mailboxes capable of sending and receiving. The feature automatically handles threading and fires webhooks on …
A developer building an email agent using Nylas Agent Accounts details the receive-think-respond loop, highlighting edge cases such as the `message.created.truncated` webhook for bodies over 1 MB and the need to avoid re…
Microsoft announced the general availability of Copilot Cowork, an agentic system for complex, multi-tool tasks, after a preview period. Over half of the Fortune 500 and companies like Accenture and Zurich Insurance are …
A developer shares techniques for improving their use of Claude Code, an AI coding tool, in a new article on Towards AI.
NVIDIA announced the ACE Game Agent SDK and new Unreal Engine 5 plugins at Unreal Fest 2026, enabling developers to build on-device AI companions for games. The SDK provides agentic, chat, and RAG APIs for creating NPCs …
Qualcomm unveiled the Snapdragon Reality Elite chipset at Augmented World Expo, targeting next-generation mixed-reality headsets with up to 48 TOPS of on-device AI performance. The platform powers the XREAL Project Aura …
The Qwen team released three embodied AI models—Qwen-RobotManip, Qwen-RobotWorld, and Qwen-RobotNav—as part of Qwen-Robot-Suite, each built on a Qwen vision-language backbone and targeting different robotics problems. Ro…
Angel Armendariz, a 20-year enterprise sales veteran from AWS, released an open-source Claude Skill called Forward Deployed Selling (FDS) that encodes a sales methodology for the AI era. The skill, free and installable i…
Canadian LLM developer Cohere reports a surge in inbound interest after the US blocked Anthropic's latest models, positioning itself as a sovereign alternative to US AI providers.
SpaceX will acquire AI coding tool Cursor for $60 billion in an all-stock transaction, the companies announced today. The deal comes days after SpaceX's IPO and months after its merger with xAI, as Cursor struggles with …
George and Ann launched Memento, an open-source, self-hosted tool that transforms email archives into a searchable personal wiki using agentic search and LLMs. It organizes emails into People, Projects, Concepts, and New…
Dino, an AI coding agent designed for collaborative development, launched on Hacker News. The tool emphasizes transparent reasoning and keeps developers involved in the coding process rather than operating autonomously.
A new benchmark, Snyk VulnBench JavaScript 1.0, reveals that large language models (LLMs) produce inconsistent vulnerability findings across repeated scans of the same JavaScript code, with 80 of 161 unique unmatched fin…
A developer argues that hardcoding a single LLM per feature is inefficient, as it either overpays for simple requests or underperforms on hard ones. They propose difficulty-based auto-routing to send each request to the …
A developer describes a technique to reduce token waste when feeding web content to large language models in RAG pipelines. By rendering dynamic pages in a headless browser and converting the DOM to clean Markdown, token…
Pipa, an autonomous Slack agent built on OpenCode, handles operations tasks like planning, writing, research, and repo cleanup without requiring a developer's laptop or terminal. The architecture uses Slack for communica…
Continual learning (CL) for LLMs remains largely undeveloped, but researchers are exploring three categories of attack to make it safer: interpretability, controllability, and evaluation. The authors caution that many pr…
Mistral has adopted a plump cartoon cat as its mascot on the Vibe landing page, directly referencing the viral 'Le Chaton Fat' AI meme that briefly fooled researchers into thinking it was a real model. CEO Arthur Mensch …
Microsoft's Copilot Cowork, a persistent multi-step AI agent feature for Microsoft 365, is now generally available worldwide. The system integrates Anthropic's Claude models alongside OpenAI's to autonomously execute com…