A backend platform built for AI agents, NVIDIA’s open-source conversational AI, a native macOS GUI for Claude Code, and developer tools for webhooks and mock APIs.
Massive day. Alibaba drops Qwen 3.5 on Lunar New Year’s Eve—a 397B MoE rivaling GPT-5.2 under an Apache 2.0 license. Meanwhile, the Pentagon-Anthropic feud escalates to threats of “supply chain risk” designation.
Alibaba released Qwen 3.5-397B-A17B on Lunar New Year’s Eve—a 397B MoE model with only 17B active parameters. Benchmarks put it on par with GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro across coding (83.6 LiveCodeBench), math (91.3 AIME26), and reasoning tasks.
Open-weight under Apache 2.0. Runs on consumer hardware at the 17B active parameter size. The open-source community now has a frontier-class model to run locally.
Quick hits — hardware leaks, AI philosophy, and small model breakthroughs.
Claude Code meets ComfyUI with a new MCP integration, a desktop asset manager for generations, and ByteDance faces legal heat over Seedance training data.
A local-first macOS AI assistant, Rosetta 2 deprecation warning, self-hosted medical records, and a comprehensive Discord alternatives list.
Quick hits — macOS utilities and self-hosted tools.