You don't have to say it the right way

Jude Mercer, SEO and Content, Memoe — 13 October 2026

A voice tool built around a fixed list of commands only works once you've learned the list. Ask for the wrong word, or phrase it the way you'd naturally say it instead of the way the tool expects, and nothing happens — or the wrong thing does.

• Adam Engst, the founder of TidBITS, used macOS Voice Control one-handed with a broken wrist and found it "doesn't work reliably across all apps. For instance, I can't get it to capitalize a specific word in Slack, although I can select some text and say 'Capitalize that.'" (tidbits.com, 21 September 2026)
• Commenting on that same piece, the writer Jeff Hecht put the general problem plainly: "designers fail to allow for temporary limitations of the user. The lack of an effective voice editing feature can be a serious limitation." (same piece, 21 September 2026)
• Writing a design document for voice control of a desktop app, the engineer Viktor Farcic set the bar at the opposite of a command list: built so "the user never has to guess the phrasing," with what you say resolved against what the app can actually do, not matched against a fixed grammar. (GitHub, 1 September 2026)
• On Hacker News, jiehong wished macOS "had amazing computer use and was able to control all aspects of all apps, and you'd be able to instantly do anything with voice only" — not a specific command, anything. (Hacker News, 21 September 2026)
• And more recently, crazygringo described wanting a voice tool split in two: plain dictation for text, and separately "issuing commands for a local LLM to interpret" — voice understood and acted on, not just typed out. (Hacker News, 11 October 2026)

Five different people, five different years' worth of frustration and hope, pointing at the same seam: a command list is a second thing to learn on top of the app you already know how to use.

What Moe does instead

Ask Moe to do something in an app, and it reads that app's own menu bar — the real items, in the words the app itself uses for them, along with their shortcuts — and invokes the one it needs by name. There is no separate list to learn, because the list it works from is the one already built into the app. If the thing you asked for doesn't exist in that app, Moe says so plainly, rather than silently doing nothing or guessing at the nearest match.

Screen-driving — finding a button and clicking it, the way a person would — is the last thing Moe tries, not the first. The order it works through is: the data it already has, then the app's own scripting, then a menu item or keyboard shortcut, then its own browser for anything on the web, and only then a click on the screen. That also means it can reach into apps whose main window is one solid canvas it cannot read at all — music software, CAD tools, games — because their menus are still ordinary text underneath.

It works this way across whatever app is open, from any app, by holding one key and saying what you want in your own words. Moe does the work itself; touching your trackpad stops it.

Where this still needs you

Menus only cover what the app itself exposes as a command. A setting buried in a preferences pane with no menu equivalent, or a one-off visual judgment call ("does this photo look overexposed"), still needs Moe to look at the screen — or needs you. Reading a menu by name is the fast, reliable path when the thing you want is already one of the app's own commands; it is not a way around genuinely visual or free-form work.