Keeping an idea journal when you're building a company
Not a notes app and not a task list. A record of what you decided and why, which turns out to be the thing you cannot reconstruct later.
There is real research on why insight clusters at the edge of sleep, and a practical problem it creates: the idea is useless if you cannot find it in March.
Everyone who makes things knows the pattern. The problem you pushed at all day gives nothing. Then you are half asleep, or in the shower, or walking somewhere without your phone, and the answer arrives fully formed.
This is not folklore. There is now decent evidence for what is happening, and the mechanism suggests something useful about how to catch the output.
Thomas Edison reportedly napped holding steel balls over metal pans, so that as he drifted off his hands relaxed, the balls fell, the clatter woke him, and he caught whatever his mind had been doing in the moments before.
In 2021 a team at the Paris Brain Institute tested essentially this. In Sleep onset is a creative sweet spot, published in Science Advances, 103 participants were given maths problems that all shared one hidden shortcut, which nobody was told about. Between attempts they took a short break holding an object, Edison-style.
Participants who spent at least fifteen seconds in N1 — the shallow drifting stage right at sleep onset — were about three times more likely to discover the hidden rule than those who stayed awake. Dropping into deeper sleep removed the effect. The benefit sat in a narrow window on the edge of consciousness, associated with a particular mix of alpha and delta activity.
Fifteen seconds. Three times. The effect is specific, and it is not about being rested.
The plausible account is about constraint. Directed attention is good at following a path and bad at leaving one. Fully asleep, nothing is steerable. In between, associations loosen while enough coherence remains to notice when something fits.
Which explains why the shower and the walk work too. They share the same property: occupied enough not to steer, awake enough to notice.
It also explains the thing that is genuinely annoying about it — you cannot make it happen on demand, and the state that produces the idea is the state least able to write it down.
Here is where the advice usually goes wrong. "Keep a notebook by the bed" is fine, and it solves the smaller half.
The larger half: you captured it, and in March you cannot find it.
Every maker has a graveyard of good thinking — notes apps, voice memos, a file called ideas.txt, messages sent to themselves at 3am. The capture worked. The retrieval never happened. The idea was not lost at the moment of having it; it was lost over the following six months, in a place you could technically have looked.
This is worse than it sounds, because the ideas you most need are the ones you cannot search for. You do not remember writing it. You would not guess the words you used — at 2am you wrote something elliptical that made complete sense at the time. And the moment it becomes relevant is precisely the moment you have no reason to go looking.
Three rules, in order of how much difference they make.
One place. The single biggest determinant of whether an idea survives is whether it is somewhere you will look. Four capture tools is the same as none, because nothing can be read across. Pick the one you will actually reach for half asleep and put everything there, including the ones that do not feel important.
Write the context, not just the idea. At 2am, "two-sided sync" is a complete thought. In March it is three words with no referent. One extra sentence — what problem this was for, what prompted it — is the difference between a note you can use and a note you will delete. Thirty seconds, and it is the whole game.
Date everything. The date is the only index that is always available and never wrong. Months later, "that thing around the launch" is a findable query if the dates are intact and an impossible one if they are not.
Even done well, retrieval is the unsolved part.
The idea you need in March was written in September, in different words, about a problem you then described differently. Keyword search does not connect them. Tags do not, because you could not have named it. Chronology does not, because you do not know when you wrote it.
What does work is matching on meaning rather than wording — asking for everything about a subject and getting back the entry that never mentioned it by name. That is the particular retrieval problem Ember is built around, and it is the reason an idea journal is worth more than an idea file.
There is also a second-order effect that only shows up in aggregate. The same idea recurring four times across a year, in different forms, is a much stronger signal than any one instance. You cannot notice that in the moment — each time it reads as a new thought. Seeing the four together is what tells you it is not.
You cannot summon insight, but you can make the conditions more frequent.
Load the problem first. Both the research and the anecdotal pattern depend on having genuinely engaged with something before the break. The drifting mind only recombines what is already loaded.
Then stop, properly. A walk without a podcast, a shower, lying down in the afternoon. The common ingredient is occupied-but-unsteered.
And pay attention to the edges of sleep specifically. That window is narrow — and the fifteen-second figure is a good reminder that it is far shorter than it feels from inside.
The 2am idea is not better because it arrived at 2am. It is better because it came from a mode of thinking you cannot access at your desk.
That makes it worth catching. But catching is the easy half, and the half everyone optimises. The ideas you lose are not the ones you failed to write down — they are the ones you wrote down perfectly and never saw again. Fixing the finding is the part that pays.
Not a notes app and not a task list. A record of what you decided and why, which turns out to be the thing you cannot reconstruct later.
Day One exports in three formats and only two of them are useful outside Day One. Here is which to pick, how to get it, and what survives the move.