Say it the moment you wake, before it fades. Then let the morning check who else was asleep at that hour.
No account and no questions first. Speak in any order, as broken as it comes. Time is asked after, as an estimate, and the words stay exactly as you said them.
Every night is a shape. Mornings without a dream stay as outlines, because they are half of any month. A shape that comes back is threaded to the night it came from.
Example month, from our own test dreams
When two people record nights that overlap, their reports are compared. The comparison looks for one scene seen from two sides, and it is measured only against pairs of dreams from unrelated nights.
Example, from our own test dreams
Two dreams recorded in overlapping sleep windows appear to describe one scene from different sides.
Stronger than every one of the 18 control comparisons in that run.
The other dream stays closed until both people agree to open it. Live numbers are on the machinery page, and today they are zeros: the experiment has not started.
No account needed
The seconds after waking belong to the dream. Recording works with no sign-in, and the dream comes back to you in this browser.
Opened by both, or not at all
A match never reveals anything. Two people agree first, and each sees their own words before the other does.
Your voice, your call
Recordings are kept for two years to make transcription better. One switch turns that off, one action erases them.
Step one
On waking, before it fades. Speak it or type it, in any order, as broken as it comes.
Step two
We fix the window you were asleep in, and only then look for people whose window overlapped yours.
Step three
A model checks whether the two describe one scene from different sides. It is asked first what contradicts that, and only then what matches.
Step four
The same comparison runs on dreams from unrelated nights. A finding is only worth something next to that background.
Two people can describe one scene because dreams share common material: halls, stairs, water, being late. That is why every finding is measured against pairs of dreams from different nights, run through exactly the same comparison.
If overlapping sleep produces no more matches than unrelated nights, the hypothesis is wrong, and we will publish that with the same weight as the opposite.
Almost every morning we find nothing. We say so plainly, and that answer counts as a result.
Free, and it stays free. An empty morning counts too: half of all mornings pass without a dream, and the experiment needs them.