GTD for a calendar you don't control

GTD is Capture, Clarify, Organize, Reflect, Engage. Each step has a specific failure mode when your inputs arrive as speech instead of as email.

Capture is the load-bearing one. The other four degrade gracefully; capture fails silently. A commitment that was never captured does not show up as a gap in your system, it shows up three weeks later as a surprise.

Allen's rule is that capture must be frictionless enough that you never weigh whether something is worth writing down. In a meeting there is friction he never had to account for: capturing costs you the conversation. Typing while someone explains a tradeoff means you get the sentence and miss the reasoning, and the reasoning is usually the part you needed.

The way out is to stop treating capture as one act. Split it:

• Passive capture is a complete record made without your attention — the transcript. It is exhaustive, unfiltered, and requires nothing of you during the meeting.
• Active capture is the handful of things you noticed as a commitment: a decision, something you agreed to own, a name to follow up with. Three or four lines, not a transcript.

Done this way, active capture becomes cheap enough to do honestly. You are no longer trying to write everything down, because the everything already exists elsewhere; you are only flagging what your judgment says matters. That is exactly the "mind like water" state Allen is after, and it is unreachable while you are also functioning as a stenographer.

Clarify asks two questions of every captured item: is it actionable, and if so what is the very next physical action? Meeting notes resist both, because meetings produce statements rather than actions. "We should probably revisit the pricing tiers" is not a task, and copying it into a to-do list produces a list of things that never get done.

The conversion has three parts, and skipping any one of them is what makes an action item rot:

• A verb and an object. "Revisit pricing" becomes "draft three tier options for the pricing review." If you cannot start it from the sentence alone, it is not clarified.
• A named owner. Not a team. A person. An action item owned by "marketing" is owned by nobody, which is the single most common way a decided thing fails to happen.
• The reasoning, kept next to it. Six weeks later the useful part is why the group ruled out the other two options. This is the piece pure task capture always loses and a transcript always keeps.

Do this in the ten minutes after the meeting, not in the weekly review. Clarifying is fast while you still remember the room and slow once you don't, which is the actual argument for post-meeting buffers.

Allen's contexts (@calls, @computer, @errands) were designed for a world where the constraint was location and equipment. When your constraint is other people's calendars, the contexts that earn their keep are different. Four cover most of it:

• @next — actions you own that can be started now. The list you actually work from.
• @waiting — commitments other people made to you, with the date and the meeting they made them in. This list is the whole reason GTD survives contact with a meeting-heavy role: nothing else tracks the things you delegated in a room and would otherwise chase from memory.
• @meeting — things to raise, filed per person or per recurring meeting rather than in one pile. When the 1:1 arrives, the agenda has assembled itself over two weeks instead of being invented in the corridor.
• @someday — the ideas raised and not chosen. They stay out of the active lists without being lost, and they are worth re-reading quarterly.

The @meeting list is the highest-yield addition, and the one most people skip. It converts every "I should mention this to Sam" from an interrupting thought into a filed one, which is the same trick GTD plays everywhere else.

The weekly review is the step that keeps the system trusted, and the one that gets dropped first. For a meeting-heavy role it needs a specific agenda, because the raw material is a week of conversations rather than a week of paperwork. Thirty to forty-five minutes, ideally in the lowest-energy slot of the week:

• Walk the week's calendar meeting by meeting. Every one either produced a captured decision or action, or it produced nothing — and a meeting that produced nothing twice in a row is a finding, not an accident.
• Clear the @waiting list: what has come back, what is overdue, what needs a nudge before it becomes a surprise.
• Check next week's calendar against the @meeting lists, so preparation happens against a list rather than against panic.
• Note recurring meetings that keep producing nothing. Feed those into the meeting audit rather than tolerating them for another quarter.

Allen's engage step assumes you choose by context, time available, energy, and priority. Three of those four are severely constrained on a meeting day, which changes what a good list looks like. Actions have to be sized to the gaps that actually exist: a next action you cannot start in the 20 minutes before the next call will be skipped every time until it becomes urgent, at which point it is done badly. Slicing "write the spec" into "draft the spec's problem statement" is not busywork, it is what makes the list usable on the days you have.

The complement is protecting a small number of gaps that are big enough for the actions that cannot be sliced — the mechanics of that are in protecting focus between meetings.

There is a clean mapping between Allen's inbox and a day of transcripts: both are unprocessed, both are complete, and both exist to be emptied rather than read. The transcript is the raw inbox; the @next, @waiting and @meeting lists are what remains after clarifying. What it removes is the anxiety Allen identifies as the actual cost of an untrusted system — the background hum of wondering what you agreed to and forgot.

Earkeep is built to be that inbox. It records mic and system audio continuously while it runs and transcribes locally on your device, so the record exists before you decide you need it — no bot to invite, no start/stop to remember. Meetings are defined afterwards by selecting a span on a timeline, or automatically from a connected calendar. Audio is processed only in memory and never written to disk, and each day is a plain file you can grep, so the weekly review is a search rather than an archaeology exercise. See meetings after the fact for how the retroactive spans work.

Clarifying can be delegated too: an agent reading the day's transcript can propose the decisions and action items, leaving you to correct a draft instead of producing one. That is agents and MCP, and the full loop from meeting to output is in meeting workflows.

Download Earkeep free for macOS and stop carrying meeting commitments in your head between reviews.