The theme today is infrastructure. GitHub trending is dominated by tools that treat AI agents as first-class citizens — design systems, browser automation, home integrations — while Anthropic ships two new models and starts the clock on its long-anticipated IPO. The community is no longer just building things with agents; it is building the scaffolding that agents will run on.
Anthropic Introduces Claude Fable 5.2 and Mythos 5.2
Anthropic released two new models on November 9th: Claude Fable 5.2 and Claude Mythos 5.2. Both are positioned for demanding technical work — coding, quantitative finance, and scientific reasoning. The dual-model announcement follows a pattern of offering differentiated tiers, with one model optimized for speed and one for depth, giving developers more room to match cost and capability to their workload.
Anthropic Begins IPO Roadshow
One day after the model launch, Anthropic kicked off a roadshow for a proposed IPO of Class A common stock. After years of rapid growth and massive private fundraising, the company is now making its case to public market investors. How it prices and trades will be one of the more closely watched market events in AI — a signal of how Wall Street values the category at a time when model releases are accelerating.
Meta Astryx: A Design System Built for AI Agents
meta/astryx is a customizable, open-source design system built specifically for use with AI agents — not just for humans building UI, but for agents that need consistent, structured component surfaces to interact with. It gained over 3,100 stars in a single day, the most of any trending repo today. As more interfaces are operated by agents rather than users, the design tools will need to evolve with them.
Anthropic Skills: A Public Library of Agent Capabilities
anthropics/skills is a public repository of reusable agent skills — structured, composable behaviors that extend what an agent knows how to do. It passed 192,000 total stars and added nearly 1,700 more today. If design systems help agents interact with UI, skill libraries like this help them act reliably in the world. The repo's growth is a window into how much demand there is for shareable, production-tested agent behaviors.
Home Assistant MCP: Natural Language for Smart Homes
homeassistant-ai/ha-mcp connects Claude, ChatGPT, and other AI assistants to Home Assistant via the Model Context Protocol, enabling natural-language automation and configuration of smart home devices. It added nearly 800 stars today. MCP is quietly becoming the integration layer that lets AI assistants reach into local systems, not just cloud APIs — and home automation is proving to be one of the most compelling early use cases.
github.com/homeassistant-ai/ha-mcp
Vercel Agent Browser: CLI-First Browser Automation in Rust
vercel-labs/agent-browser is a browser automation CLI written in Rust, designed for AI agents rather than test suites. It added 780 stars today. Rust's reliability and cross-platform portability make it a good fit for tooling that needs to run inside automated pipelines, and the fact that Vercel is investing here signals that browser automation is becoming a first-class concern in modern deployment infrastructure.