A model, its evidence, and what it cannot settle
Every claim here carries an evidence level and a source, every gap is published, and the whole registry is downloadable. Read the model rather than looking out from a seat inside it.
The same registry, read from where you sit
- 96
- 51
- 2
These numbers are identical on all three of these pages, because there is one registry. What changes between them is the order entries come in, the question they are put to, and which seat the lens starts in. What does not change is the entries, the evidence behind them, or the wording of a single claim.
Where you sit
You do not sit in this funnel. That is the difference this page is built on: the other two readings look out from a point inside the process, and yours looks at the process. No lens is set, so every seat stays visible at once.
One entry, two readings
This is the trace more mechanisms produce than any other, which is why it is the example. Below is what the registry records the two seats seeing — the same entry, the same evidence, written down separately because they are not the same experience.
Observation obs.generic_closer_alignment_rejection_template Generic "closer alignment" rejection template
A message sent at 03:14 with no sentence specific to this candidate, this role or this conversation, and a follow-up that returns the same wording.
The text is the standard wording for this outcome. Which requirement was the gap is not stated in it, and asking returned the template again.
A rejection step in the queue and, afterwards, a request for one line of specifics.
The wording is the one approved for this outcome; what the panel scored beyond a recommendation is not in the recruiter's record either.
Where to start
- Graph
The whole registry as a graph: entries, relations, and what reaches what.
- Data
Coverage, unplaced traces, and the mechanisms no observation can pin down.
- Methodology
Seven evidence levels, what each one licenses, and what disqualifies a source.
What this cannot tell you
Coverage is uneven and stated. Three entries carry no source because none exists that would bear the weight, the mechanisms no observation can pin down are listed, and the whole of it is on one page.
The atlas does not have what happened to you
Then it is incomplete, and this is the most useful thing anyone can tell it. Write what you received — the message, the silence, the change — in your own words. Nothing sent here appears on the site: it is read by a person, and only a pattern that several people describe becomes an entry.
