Nothing publishes itself
The Human Seal.
A person approves every situation before it reaches anybody, and approves every reading before it is published. Neither happens automatically, on a schedule, or because a threshold was reached. The approval is recorded with a date and a card version, and a card or reading without one does not exist as far as the platform is concerned.
Two moments, not one.
The seal is applied twice, at two different points, for two different reasons. Conflating them is easy and hides the thing that matters, so they are worth separating.
Before release. A drafted situation is reviewed and either approved or rejected. The reviewer has to be able to argue for every option honestly; if they cannot, the card does not ship. This is where gray against gray is actually enforced, rather than merely declared.
Before publication. Once a card crosses its response threshold, its reading does not automatically appear. A person looks at the card and its aggregate together and signs off on release. Crossing a threshold makes a reading eligible; it does not make it published.
Why a person and not a rule
Most of what makes a card good or bad cannot be checked mechanically. Whether two options are written in the same register, whether a situation reads differently to somebody in a different household, whether the honest answer would be embarrassing to give — these are judgements, and a rule that pretended to check them would be a rule that passed everything.
There is a second reason, and it is the more important one. A rule can be adjusted quietly. A person has to decide, and a decision has a name and a date attached to it. When something goes wrong, an automated pipeline offers nobody to ask; a recorded approval offers somebody who has to answer for it.
What the seal records
The date of approval and the exact card version approved. That version is carried through to any reading published from it, which is why a published reading names its version — a later reader can check that the wording answered was the wording sealed, rather than a wording tidied up afterwards.
What the seal cannot do
It cannot make a figure accurate, cannot rescue a badly written card once people have answered it, and cannot remove a reading that turned out unflattering. Approval runs forwards only. A card that should not have shipped stays in the record along with what it collected, because a record you can retroactively unapprove is not a record.
Somebody has to be answerable. A pipeline that approves its own output leaves nobody to ask when a card turns out to have been a mistake.
Before release
What the reviewer
actually checks.
Six questions, asked of every card. A no to any one of them sends the card back rather than through.
| The question | What a failure looks like |
|---|---|
| Can every option be argued for honestly? | One option can only be defended by pretending |
| Are the options written in the same register? | One is warm and long, another curt — a right answer in disguise |
| Does somebody real pay, whichever way you go? | One option harms a person, another harms nobody |
| Is the honest answer safe to give? | Answering truthfully would be embarrassing, so people will not |
| Is it settled by law or by a fact? | The answer can be looked up, which makes it a quiz |
| Is it outside politics and religion? | Answers would become declarations of identity, not accounts of conduct |
Where these tests come from. Gray vs gray
Before publication
Eligible is not published.
A card crossing its threshold becomes readable to the people who answered it. Publishing that reading on the public record is a separate act, and it needs a separate approval.
The threshold does not publish
Reaching enough responses makes a reading eligible. Somebody still has to look at the card and the aggregate together and decide it is fit to stand on the record.
No schedule forces it
There is no weekly slot to fill and no target number of published readings. A month with nothing sealed is a normal month, not a failure to be corrected.
Approval is not editorial
The reviewer is not deciding whether a result is interesting or flattering. They are checking that the card was sound and the aggregate is what it claims to be.
It only runs forwards
There is no unsealing. A published reading cannot be withdrawn because it became inconvenient, and nothing about it is edited afterwards.
In one line
The definition.
- The Human Seal
- The recorded approval by a person, required twice: before a situation is released to anybody, and before a reading is published on the public record. It carries a date and the exact card version approved. Nothing on Democrez publishes automatically, on a schedule, or because a threshold was reached.
Where you can see it
- On every published reading. Seal date and card version are two of the fields a reading must carry before it can appear at all. The full list
- In what is missing. The public readings page is empty, and not because cards are unavailable — because none has been through both gates yet.
Questions
Straight answers.
Who is the person applying the seal?
Someone on the Democrez team with the authority to reject a card outright. What matters structurally is that it is a named individual whose approval is recorded, rather than a process that approves its own output — because when a card turns out to have been a mistake, there has to be somebody to ask.
Does AI approve anything?
No. AI drafts situations and proposes wordings, and that work all happens before review. Approval is a human act at both points — before release and before publication — and it is the boundary that AI does not cross on this platform.
Can a card be published without a seal?
No. A card without an approval cannot be released, and a reading without one cannot be published. Crossing a response threshold makes a reading eligible; it does not make it public.
Can a seal be withdrawn later?
No. Approval runs forwards only. A card that in hindsight should not have shipped stays in the record along with everything it collected, because a record that can be retroactively unapproved is not a record of anything.
Does the reviewer decide whether a result is interesting?
No, and this is a deliberate limit. They check that the card was sound and that the aggregate is what it claims to be. A reviewer choosing which findings deserve publication would be an editor, and the record would become a selection rather than a record.
What happens to a card that fails review?
It does not enter the deck. Most failures are about writing rather than subject — options in different registers, or a situation that reads as having an approved answer — and those go back for redrafting rather than being discarded.
Does this apply to institution cards too?
Yes. Situations written for a particular college or organisation go through the same review before a session runs. What differs is where the responses go afterwards: institution data feeds that institution's report and never enters the public record.
Every card you meet passed this.
One situation a day, each one approved by a person before it reached you.