The Guardian on Meta's new Muse assistant, a cute "little guy" that answers emails, books travel and does your food shop — and why its design leans on cuteness.
Why I recommend it: Free to read. Worth reading alongside the Mouse post on Muse's file access: a friendly face makes it easier to forget how much access an assistant has.
A free one-page PDF by James Maduk, dated 21 September 2026, on how to brief an AI assistant so it produces something usable. Gives an "Outcome Brief" of five questions to answer before handing over a job (what you want to make possible, who it is for and what is broken, what "done" looks like, what is off-limits or needs your approval, and the smallest version you could test), a three-question plan check for afterwards, and three worked before-and-after examples for client reviews, social posts and client onboarding.
Why I recommend it: The most useful line on it is the off-limits question — deciding in advance what never goes out without your sign-off is what keeps this safe for client work. Print it and keep it beside you the first few times; it turns "help me with client reviews" into a request a tool can actually deliver.