How to use AI during a meeting without anyone knowing (Mac, 2026)

The recap economy assumes the useful moment is tomorrow. It is not. The useful moment is the eight seconds after a director asks what your Q3 churn number was. It is the thirty seconds a candidate has to answer a system design question. It is the minute a client spends explaining a contract clause you have never read. A transcript at 5pm does nothing for any of those.

ChatGPT in a browser tab solves part of this. You can paste a question, get an answer, and read it back. The problem is the tab. You have to alt-tab away from the call, type the question with your hands visibly off-screen, wait for tokens to stream, then read in a slightly different posture than you had two seconds ago. Anyone watching your face knows. Anyone watching your shared screen sees the tab. The trick is not having access to a model. The trick is having access without breaking eye contact.

Whisply was built for that gap. The overlay sits on top of whatever is on screen, summoned with one keystroke, answered in plain text inside the same field of view as the meeting window. You read it the way you would read a sticky note on your monitor, except the sticky note knows what was just said and updates as the conversation moves.

macOS exposes a window-level flag for content protection. When it is set, the compositor refuses to include that window in any frame it hands to a screen recorder, a screen share, or a system screenshot. Whisply ships with that flag on for every overlay window by default. The technical detail is boring. The consequence is not. You can share your entire display in Zoom and the person on the other end will not see the Whisply window, the answer text, or the cursor blinking inside it.

This is not a hack and it is not a race condition. It is the same mechanism Apple uses to keep DRM-protected video out of screenshots. Whisply uses the public API. The same protection applies to QuickTime, OBS, Loom, and any third-party recorder that goes through the standard macOS capture pipeline. The overlay is private to your local display, full stop.

Audio is one-way. Whisply listens to your microphone and to system audio (so it hears the other side of the call as well as you), but it never injects audio back into Zoom, Meet, or Teams. There is nothing for the meeting platform to pick up. The other participants hear you, and only you.

Whisply lives in the menu bar. You open the meeting in Zoom or Google Meet or Microsoft Teams the normal way, you join, and you do nothing different. When you want help, you press Cmd+Return. A compact text field appears in the top portion of your screen. You type a question (or you say nothing and let Whisply answer based on what it just heard), and the answer streams into the same field within a second or two.

Cmd+Return again hides the overlay. Cmd+Shift+Return cycles through recent answers if you missed the first one. Cmd+K clears the conversation. The hotkeys are designed so you can drive the entire interaction without your hands leaving the home row, and without your eyes leaving the meeting window. The motion looks like you glanced at a notification and went back to listening.

If the host asks you to share your screen mid-conversation, you do not need to close anything. Hit share. Whisply is still on top of your display, still updating, still invisible to the share. The person on the other end sees your slide deck or your code editor and nothing else.

Recall is the obvious one. A name you should remember, a number from last quarter, the gist of a document you skimmed on Monday. Whisply can pull from what it has heard in the current call and answer with the right context, so you do not have to say the awkward part out loud.

Phrasing is the underrated one. You know what you want to say but the sentence is not landing. Type three words into the overlay and read the cleaner version back in your own voice. The judgment is still yours. The wording is sharper than the wording you would have improvised under pressure.

Translation works in both directions. Non-native English meetings, accents you find hard to parse, technical jargon from a field next to yours. Whisply transcribes the last thirty seconds of the call into the overlay if you ask, and explains what was meant in language you actually use. It is the difference between nodding and following.

For interviews, Whisply can hold a structured brief in the background (the role, the company, the things you decided to emphasize) and surface a relevant talking point when the interviewer pivots. You stop hunting through your prep notes in another window. The prep is just there when you reach for it.

Zoom sees you. It sees your camera feed, your microphone, the application windows you chose to share, and nothing else. There is no Whisply participant. There is no Whisply bot in the lobby. There is no second audio track. The Zoom participant count is one for you, the same as it would be without any AI tool installed.

Google Meet, Microsoft Teams, Webex, and FaceTime behave the same way. Whisply does not dial in. It does not impersonate a user. It does not request a join link. If your company has a policy against bots in calls (and most do, increasingly), Whisply does not trip that policy because there is no bot to detect.

Recordings made by the meeting platform contain no trace of the overlay. The cloud recording that gets emailed to the host shows your face, your screen share, and the audio of everyone who spoke. The overlay is local to your machine and stays there.