ALargely tested
Epistemic
Can the instrument distinguish what evidence establishes from what it does not.
Where this is
Find which unknowable fact has a person attached to it.
Semafide has finished enough internal work to know what it can and cannot establish. The remaining bottleneck is external. Standing cannot be manufactured in the repository.
BNext
Economic
Does somebody currently incur enough cost from one of those facts that they will pay an independent party for a forward record.
CLater
Institutional
Will other institutions rely on Semafide’s statements in consequential decisions. That is standing. It is not yet testable at full scale.
Destination
A party whose statements about automated executions other institutions rely upon.
Not the owner of a log. Not a cryptography company. Not an open-source verifier company. Not a diligence service. Those may be mechanisms. Standing is granted only when another institution changes its behaviour because it accepts something Semafide said.
A useful private tool may be the path toward standing. It is not standing itself.
Two facts
They both sound like “we cannot reconstruct what happened.” They fail for different reasons.
Pick the object this next conversation is actually testing. Both is two conversations.
Forward only
Last year’s file cannot be reconstructed.
Overwritten cascade settings are gone. Runs that were never sealed are gone. You are not selling reconstruction of a past valuation. You are selling forward preservation, against a cost they already understand from the past.
The first product sentence, on either object: from now on, this fact exists at execution, held by someone who does not produce the valuation.
The fork
Would they pay to know, for themselves, if nobody else ever accepted the statement.
Utility-first
One buyer, own exposure
An institution already bears a cost from not knowing. They will pay to reduce that uncertainty for themselves. The first output is an independent determination they consider useful enough to act on. No examiner required. Standing may accumulate later from usefulness. This is the easier entry, if it exists.
Acceptance-first
Worthless until honored
The information has little value unless an examiner, a court, an insurer, or another institution gives it formal effect. Then you are trying to found an evidentiary institution before there is a bilateral product. That is the harder company.
“People think this is interesting,” “a regulator might someday require it,” and “a court could find this useful” are all capable of a false positive. Someone can agree the problem is real and never pay for an answer.
Who counts
Interest is not a design partner.
| Role | What they did | What they are not yet |
|---|---|---|
| Interested person | Agrees the problem is real. May praise the instrument. | A buyer. A capture seat. |
| Design partner | Names a cost they already eat, picks one object, can furnish a capture seat for that object, will use a forward record internally. | Standing. A certificate. |
| First buyer | Pays for their own use. Utility-first. | An institution that honors the statement for someone else. |
The conversation
Walk in with one question. Walk out with a kill or a partner.
Do not lead with the repository, the map, or a pitch that the market “needs evidence infrastructure.” Take the question. If they have a technical person later, the repository is the practice statement that person can attack.
Ask this
The last time you could not reconstruct an automated valuation — was it that you did not know how many runs were in the file, or that the settings that produced the number no longer existed? Who paid. Would you pay, going forward, to have that one fact held outside the party that produced it, even if no examiner ever accepted the statement?
Reading
No conversation recorded yet. The next proof is a person, not another pass on the instrument.
Stored in this browser only. Download HTML or Markdown to keep a copy, including this conversation.
End to end
The path, in order, so the work does not loop.
- 01Claim-discipline on the instrument is done enough. Stop using the repository as a substitute for contact.
- 02Choose one object for the next conversation. Completeness or configuration. Not both.
- 03Find a person who already eats the cost when that fact is missing — repurchase, QC, warranty, a desk that had to reconstruct a file.
- 04Ask the question. Copy it. Do not expand it into a market narrative in the room.
- 05Read the kill. No cost named: that object is dead here. Config without a digest: dead. Count without a capture seat: not a partner yet.
- 06If they will pay for their own use and can furnish a seat: they are a design-partner candidate. First engagement is forward capture of that one fact, held outside the producer.
- 07What they get: an independent determination they can act on internally. What they do not get: reconstruction of old files, a compliance certificate, or standing.
- 08After one useful private loop you have utility, not standing. Standing is when a second institution acts on a Semafide statement.
Stopping rules
What this brief is for, and what to stop doing.
- The repository is a practice statement. Harden it until an obvious technical reviewer no longer has an easy reason to dismiss the thesis. Then stop. It does not answer whether anyone will pay.
- The public site is a front door, done enough. Another homepage pass is not a design partner.
- The map is for a reader who already opened the repo. It is not the pitch.
- This brief is the working instrument for Proposition B. Fill the conversation. Let it kill an object. That is progress.