There was unauthorized access to Anthropic's Mythos, and some users used it even before the official testing began.
It has been alleged that a promotional image for OpenAI's Images 2.0 is identical to an existing photograph.
Tesseron released, controlling running apps via a Claude Code plugin.
The Graphify skill for Claude Code turns codebases into a knowledge graph and claims to greatly reduce token usage.
It's an MCP server that lets Claude directly access Google Ads accounts to diagnose and optimize campaigns.
Roo Code is shutting down after surpassing 3 million installs, shifting full focus to Roomote.
IBM has released Granite-4.1-8B, enhancing tool calling and instruction-following performance.
It's a Tauri-based desktop app that lets you access local AI coding agents via messengers or a web terminal.
Claude Code has disappeared from the Claude Pro pricing page's list of included features.
A 4-bit reasoning model based on DeepSeek-R1-32B has been released, reduced to 20GB.
Meta is using mouse, keyboard, and screen data from US employees' PCs to train AI agents.
Unsloth has released a GGUF distribution of Kimi K2.6 along with related documentation.
Replit has been named a 2026 Google Cloud AI Tooling Partner.
Mozilla found and fixed 271 Firefox vulnerabilities using Anthropic Mythos Preview.
Sergey Brin joins Google's special team to lead the response to Anthropic/Claude.
It has been reported that the NSA is using Anthropic's Mythos preview model.
Even with 32GB VRAM, llama.cpp's `--fit` ran Qwen3.6 Q8 at a 256k context at 57 t/s.
Replit unveiled Security Agent, which reviews an entire codebase in under 1 hour.
Cal.com has released a self-hostable MIT-licensed fork with commercial code removed.
Running the same prompt across 9 local models revealed a correlation between quantization provider differences and code length.
Meta collects mouse and keystroke input from US employees' PCs and uses it to train AI models.
Meta has decided to collect employees' mouse movements and keystrokes to train AI models.