A GTD inbox is a decision queue. Things land in it unsorted and half-formed, and the only work you do there is deciding what each one is — never the work itself. That distinction is what separates an inbox that empties in eleven minutes from one that quietly turns into a second task list nobody wants to open.
The failure is easy to spot from the outside. If items sit in your inbox for more than a few days, you are not capturing badly. You are processing badly, usually because processing has silently become doing.
What a GTD inbox holds
Anything you have not decided about yet. A voice note, a screenshot of a train time, a link someone sent, a receipt, "ask about the roof". No structure, no categories, no tags. Structure is what processing adds; demanding it at capture time slows capture down, and capture that is slower than the thought does not happen.
David Allen's five-step loop names the steps capture, clarify, organize, reflect and engage, and the inbox lives on the seam between the first two. Why clarify has to be a separate step from capture at all — and why apps that force filing at entry break the method — is argued in how to pick an app for GTD. Take that as given here.
What matters for this post is that the inbox has no other job. It is not a list of things to do today, not a reading list, not a place to leave something you are undecided about. An item that is still there after two processing passes has been decided against without anyone admitting it.
Five questions, in order, with a cost
Processing is a fixed sequence. Ask the questions in this order and most items leave after the first or second, which is what makes a full inbox less frightening than it looks.
| the question | if yes | if no | seconds |
|---|---|---|---|
| is it actionable at all? | continue to the next question | reference, shelf, or delete | 3 |
| will the whole thing take under two minutes? | do it now and do not file it | continue | 2 |
| is it one step or several? | one step becomes a next action | several becomes a project plus its first step | 5 |
| is someone else doing it? | waiting-for, tagged with the date you asked | it stays yours | 3 |
| does it need to be visible this week? | today or next 7 days | the someday shelf | 4 |
Seventeen seconds if an item runs the whole gauntlet, so forty items is under twelve minutes of arithmetic worst case. In practice roughly half exit at question one as reference or deletions, so a full inbox usually takes six or seven minutes, and the panic is out of proportion to the work.
Question one is the one people get wrong, because "actionable" gets read as "important". A conference talk you want to watch is not actionable. It is reference attached to whatever it is reference for, and pretending otherwise is how a to-do list ends up with fourteen links on it.
Question three has the biggest downstream cost. Every item you leave as a project name without a first step becomes an invisible blockage that only the weekly review will find, and the rewrite it needs is the whole subject of how to write a next actions list.
Process the inbox in one pass
- Open the inbox and nothing else. No email, no calendar, no browser. Processing with the mail client open is how a twelve-minute pass becomes ninety.
- Work top to bottom in the order things arrived. Do not scan for the easy ones — reordering is a decision, and it costs the same as processing.
- Take one item and run the five questions. Give it a destination before you look at the next one.
- Do the under-two-minute things immediately, then stop. Anything that turns out to be four minutes goes back into the queue as a next action rather than getting finished out of momentum.
- When an item is reference, file it against the project it serves, not into a general notes pile. Reference with no project is the same as deleted, just slower to admit.
- When an item is wanted but not now, it goes on the someday maybe list with a review date. Not the bottom of the active list.
- Delete the last third without guilt. A third of any inbox is a thought that was interesting for ninety seconds.
- End at zero items, even on a bad pass. Zero is the state that makes the next pass cheap, because an inbox with three leftovers gets read as an inbox with three items you already failed.
Step eight is the one that decides whether this survives. An inbox that ends at four every time trains you to skim, and a skimmed inbox stops being trusted within about a fortnight.
When to process, not just how
Processing belongs in windows. Two or three fixed times a day beats a continuous trickle, and the reason is measurable rather than aesthetic.
Kushlev and Dunn's 2015 study in Computers in Human Behavior, checking email less frequently reduces stress, put 124 adults through two weeks: one week limited to checking email three times a day, one week unlimited. The limited week produced significantly lower daily stress, and lower stress predicted better well-being. That study is about email rather than a GTD inbox, so treat it as adjacent evidence rather than proof — but the mechanism it points at, a queue you visit on a schedule instead of one that visits you, is the same mechanism.
Two windows work for most people: ten minutes mid-morning and ten minutes before you stop for the day. Capture stays continuous, because capture is cheap and interruption-proof. Only the deciding gets batched.
The three items that break every pass
Some things resist all five questions. Three types account for nearly all of them.
The half-thought. "Something about the pricing page." You wrote it in a lift and the context is gone. The rule is to give it thirty seconds and then delete it — a half-thought you cannot reconstruct is not recoverable by staring, and if it matters it will arrive again.
The thing with a person attached. "Talk to Sam." These stall because the real decision is what you want from Sam, which is work, not processing. Write the sentence you would open with, put that on the actions list, and move on.
The reference with no home. A brilliant article that belongs to a project you have not started. This is the one the method handles least well, and it is exactly where GTD and PARA disagree about where material should live — the comparison is in GTD vs PARA. Until you pick a side, park it on the shelf with the project it implies written next to it.
Where init.Tasks fits
init.Tasks is one terminal-styled app holding to-dos, personal projects and a second brain, on iOS, macOS and web, built around the idea that a captured thing does not have to be a task. Anything leaving the inbox can land as a to-do, as a subproject under something bigger, or as material sitting beside the work it belongs to.
The practical effect on a processing pass is that question one stops being a trap. "Not actionable" has a destination that is not a notes app, so filing a PDF next to the project it belongs to takes the same two seconds as making a task. Quick add reads plain language — chase the roofer 9 oct #house files itself — and links fetch their own titles, which removes the small annoyance of naming things during a pass.
The weekly pass over whatever the daily passes missed is covered in the GTD weekly review checklist. The AI connection is coming in september and can do the tedious half of a backlog: point a connected assistant at a two-hundred-item inbox read-only and ask which items are duplicates. Invites go out in september and everything is free while in beta, so the waitlist is the way in for now.
FAQ
What is the GTD inbox?
The GTD inbox, called the in-tray in the original method, is a single collection point for anything you have not yet decided about. Its only rule is that everything unprocessed goes there and nothing stays. It exists so capture can be instant and thoughtless, with the thinking moved to a separate pass later.
Is a GTD inbox the same as inbox zero?
They overlap but they are not the same. Inbox zero is about email and, in its original form, about how much attention your email gets rather than how many messages are showing. A GTD inbox is a decision queue that email feeds into along with notes, screenshots, voice memos and paper. Emptying it means every item has a destination, not that every message has a reply.
How often should you process a GTD inbox?
Twice a day covers most people, plus one full pass during the weekly review. More often than that turns processing into a checking habit, which is the thing batching was meant to prevent. Less often and the queue grows past the point where a single sitting can clear it, which is when people start skimming.
What do you do with an inbox item that is not a task?
Give it one of three homes. Reference goes with the project it serves, so it is there when you actually need it. Something you want but cannot start goes on the someday shelf with a review date. Everything else gets deleted, which should be about a third of the queue, and is the outcome people resist most.