Catch first; sort later
Why the brain is the wrong store
The working memory of a person is shallow and easily overwritten. Anything held there competes with everything else trying to occupy it: the next meeting, an unresolved conversation, a half-formed worry. The cognitive cost of trying to keep ideas in mind throughout a day is itself a major drain on attention. The discipline of treating the brain as a poor warehouse and an excellent processor — moving the storage out to an external system the moment something appears — is the foundation of nearly every productivity tradition worth studying. The point is not that the system is clever. The point is that the head is free.
Capture must take seconds, not minutes
A capture system that demands organisation up front will fail under load. The right design is friction-free: a single inbox or note, accessible by keyboard shortcut or voice, that accepts unsorted input — fragments, half-sentences, questions, links, names, future tasks — and never asks the worker to file anything in the moment. The discipline of the moment is to capture as much as possible as quickly as possible. Sorting comes later, when the working session is over and the cognitive cost of categorisation is lower because the urgency has passed.
The category an assistant changes most
What an assistant changes about this loop is the sorting step. A traditional second brain leaned heavily on the user to read the inbox, decide which fragment belongs where, and file it. A modern equivalent can hand most of that decision to an assistant — read the fragment, classify it into the four or five categories that match the user's structure, propose links to existing material, and surface anything that looks like a duplicate. The user keeps the final approval but loses the friction. The traditional excuse for an empty inbox was that filing was tedious. That excuse is gone.
What to capture, in practice
A serviceable capture habit covers four kinds of input. Ideas — anything that, if revisited a month later, might turn into a project. Questions — anything you noticed you did not know. People — names, contexts, what was promised, what is owed. Tasks — anything with a defined finish line, whether the deadline is tomorrow or next quarter. The five-second test is enough. If a fragment plausibly fits any of the four, it goes in the inbox; if it later turns out not to matter, the discard is trivial. The cost of capturing too much is a brief weekly cull; the cost of capturing too little is the work that never happened because the brain forgot it could.
Think back over the last forty-eight hours. What ideas, questions, or follow-ups passed through your head that you have already lost?