Sonnet 5.5 is worth a try
Manus vs Muse and examples for computer use
TL;DR
- Anthropic released Claude Sonnet 5.5, a capable model competitive with Opus 5.5 in coding and significantly improved in image/chart understanding.
- Manus 2.0 offers a video editor, game builder, and automations, while Cue introduces consumer agents with unique digital identities.
- GPT-6 Astra is noted for its excellent computer and browser usage, surpassing other models in these tasks, with OpenAI fixing a related bug.
- OpenAI's agents have been observed interacting with Australian government websites, sparking discussions on their capabilities and security.
- Meta's new Enterprise Platform offers Muse and AI APIs to companies, led by the former CEO of MongoDB.
- Other mentions include Eleven v4 for speech generation, a subagent for HTML dashboards, a research lab approach to product teams, Clay's AI writing policy, context.dev for web answers, OpenAI's WebMCP Challenge, WYSIWYG editor building, Clicky collaborators, OpenShell for agent permissions, the ease of spotting AI-written content, UFO for agent harnesses, Fo as a personal AI, smart investing risks with agents, the impact of coding agents on software engineering, and the cost of software polish.
- AMD is acquiring World Labs for $8.2B, with Fei-Fei Li joining AMD as Chief Scientist.