what it returns
miniti reads the live transcript and returns four short, grounded sections:the model is instructed to use the transcript and leave unanswered sections empty when there is no evidence. it can still make mistakes; check important questions and decisions against the transcript.
how to open it
- macos - click the labelled catch me up control with the catch-up clock icon, or press
⌘⇧Z. the result appears in a popover attached to the button. - ios - tap the labelled catch me up control immediately left of stop. the result appears as a bottom sheet.
minutes & cost
- does not consume minutes - catch-up is part of the insights flow, not transcription
- on managed mode, the request runs via miniti’s backend using
gpt-6-luna - on byok mode, the request goes directly to your openai account using pinned
gpt-5-mini-2025-08-07
managed catch-up uses
gpt-6-luna from 2026-11-01. existing managed clients receive this change without an app update. byok keeps its current model.what’s sent
- the current meeting’s transcript (what you’ve captured so far)
- if connected, the list of calendar attendees (same hint path as speaker names)
tips
- the fresher you open it, the tighter “recent” is - it uses the tail of the transcript, not a summary
- if the transcript is too short (early in a meeting), the sections may come back empty. wait for a little more audio.
- it works best on calls where miniti can hear remote speakers (macos system audio on). on ios mic-only, it can only catch up on what you heard aloud.
related
- questions mode - suggested follow-up questions during and after a meeting
- ai insights - the same insights pipeline
- keyboard shortcuts -
⌘⇧Z
from 3.7.0: byok catch-up uses
gpt-6-luna with reasoning off. update the app and check that your openai account has model access. managed catch-up already uses luna.
