Every productivity system dies the same way: nothing breaks it, it just stops being maintained. Open loops accumulate, stale tasks rot in place, and eventually you stop trusting the list — which is the moment the system is functionally dead. One recurring review is what prevents that, and it's the highest-leverage thirty minutes in the week.
Why it's cognitively necessary
Unfinished tasks occupy disproportionate mental bandwidth. The Zeigarnik effect describes this: uncompleted items stay more accessible in memory than completed ones. This is genuinely useful when it drives you to action, and genuinely costly when there are forty of them and none are currently actionable.
A review does what your working memory can't:
- Closes loops — every item gets process, defer, or delete, which removes it from open-ended mental rotation
- Restores trust in the list — a list you believe is current gets consulted; one you don't gets ignored
- Separates planning from doing — deciding what matters is a different cognitive mode from executing, and mixing them degrades both
- Re-establishes direction — a week of reactive work drifts; the review corrects the heading
The protocol
Deliberately simple. Six passes, about thirty minutes, same slot every week.
1. Clear collections (5 min)
Gather everything loose: inbox, chat messages to yourself, notes app, screenshots, voice memos, physical paper, browser tabs you parked. Nothing gets processed yet — you're creating a single pile so that later decisions are made with everything in view.
2. Process to zero (5 min)
Each item gets one of four outcomes, decided quickly:
- Do it now — under two minutes
- Schedule it — it goes into next week
- File it — reference material, moved out of the way
- Delete it — and notice how much of the pile is this
3. Review the past week (5 min)
Look at what actually happened, not what you planned. Which commitments were met, which drifted, which meetings produced nothing. Scan the calendar day by day — it's a better memory cue than a task list.
The most valuable question here is simply why something didn't get done: too big, unclear next action, wrong priority, or actually not important. The answer determines the fix.
4. Review the projects and commitments (5 min)
Every active project gets checked for a real next physical action. Anything without a next action stalls invisibly, which is the most common failure in task systems — items that look active but aren't moving.
Then do the part people skip: kill something. Stopped projects sitting on the list consume attention without producing anything. Removing one is often more valuable than adding three.
5. Review goals and check direction (5 min)
Check against longer-timeframe goals: did this week move anything that matters? This is the only defense against a permanently busy, permanently static month. If several weeks pass with no movement on anything you claim to care about, that's information — either the goal isn't real, or nothing is protecting time for it.
6. Set up next week (5 min)
Look at the upcoming calendar and pre-decide:
- The two or three outcomes that would make the week successful
- A protected block for the most important piece of work, before the week fills up
- Anything that needs preparation before a meeting or deadline
Common mistakes
- Making it too long. An elaborate two-hour version survives two weeks. Thirty minutes is robust precisely because it's cheap.
- Planning instead of reviewing. Cleaning the list isn't strategy. Both passes are needed; the temptation is to do only the comfortable one.
- Never deleting anything. Lists that only grow become unusable.
- No fixed slot. Anchoring to a specific time and place matters more than any other variable. Friday afternoon and Sunday evening are the most common because they sit on genuine week boundaries.
- Expecting it to fix everything. The review surfaces problems; it doesn't solve capacity limits or unclear priorities elsewhere in your life.
In the AI era
AI tools can now summarize your week — your commits, documents, threads and calendar — in seconds, which makes steps 3 and 6 considerably faster. Use that for the mechanical gathering. Keep the judgment calls: what to kill, what actually matters, which single block gets protected. Those are decisions, and delegating them is how people end up efficiently working on the wrong things.
The one-line version
Once a week, thirty minutes, same slot: gather everything loose, decide on every item, look honestly at what happened, delete something, pick two or three outcomes for next week. That's it. It is the single practice most responsible for a system still existing in six months.