AI News

Meta launched Muse as a personal AI agent: something you can ask to handle tasks, organize plans, and work through projects. It’s available in the US through its app, website, and WhatsApp. Meta says everyday use is free, with paid plans for people who want to do more.

What interests me is the audience. This is aimed at people figuring out how AI fits into their day, outside a company-wide rollout. That makes the free entry point worth paying attention to.

There’s also a connection to this week’s theory. In a separate September 15 post, Zuckerberg argued that companies don’t need to wait for everyone to agree before acting responsibly. Each lab can slow down when safety requires it. He pointed to Meta delaying Muse as an example.

I’m interested in seeing whether the experience matches the pitch. Watch his interview, then try it on something small enough that you can judge the result yourself.

Latest Models Update

I’m adding Muse to the watchlist because of its consumer focus and free access. Muse is the app and agent. Muse Spark is the model Meta names underneath it. That distinction matters when comparing it with the models we discussed last issue.

Most of us still use AI one conversation at a time. Give it a task, get a result, come back when we need something else. OpenAI introduced Dots on September 29 with a different idea: an assistant that can keep working toward a goal between conversations.

OpenAI says each Dot has its own cloud computer, uses the apps you connect, and carries context across ongoing work. One example: customer feedback comes in, the Dot identifies a recurring issue, builds and tests a fix, and brings it back for review. You decide what it can access and which actions need approval.

This fits the question in this week’s working theory. Dots runs on GPT-6 Astra. The new opportunity is in how that capability is put to work. A model we already have, connected to the tools and continuity needed to take on more of a process.

The video makes the idea easy to picture. It’s OpenAI’s demonstration of the product, so I’m interested in how it holds up in everyday use. What could you hand off as an ongoing responsibility, and how would you know it was being done well?