Captured Evidence
Tier one — what the log can claim, and what it cannot
A payment arrived at 11:04:22. A thermometer reported 38.4°C. A customer clicked Submit. A judge entered a verdict. A telescope recorded a photon count.
These share one property: they entered the permanent log — tier-one captured evidence. They are not tier zero (the Real they are about). They are not tier two and above (claims ranked by support). They are substrate: the closest honest systems get to the Real, and never identical with it.
Reserve ground truth for tier zero. Tier one is captured evidence — fallible transduction every time.
The Real named tier zero. This article names what tier one can and cannot honestly claim before we ask how different captures enter the log (Four Capture Modes) or how agents emit before they commit (Thinking by Writing).
The physicist’s ladder
Physics shows the shape. Events and measurements form substrate; theories sit above and are overturned without un-happening the record.
The experiment was run; the track appeared; the instrument reported values with error bars. Newton was superseded; the old chamber photographs remain. Well-built systems mirror this: tier-one events permanent; interpretations provisional.
What is non-debatable — and what is not
Substrate claims are narrow by design. They record what entered the log — not everything we wish were true about the world.
For binding captures (signatures, sends, tombstones — typed fully in article 7): what is non-debatable is this mark, by this credential, at this moment, entered the record. Whether it expresses a free, sincere, correctly-transcribed binding — and whether the binding was wise — are claims above substrate.
For instrument captures (readings, files at boundary): this instrument reported this value under this chain. Whether the value reflects the world is a separate claim.
The series-opening failure — storing “the patient has pneumonia” as if it were the X-ray — is inference-as-fact: collapsing a claim into substrate.
The inference split still applies: the content of a score is uncertain; the act that this agent produced this output at this moment can be logged honestly as substrate when the capture mode says so — see Four Capture Modes.
Capture gaps (every mode)
Every tier-one record approximates the Real or an agent’s stream. Gaps differ by how capture happened; none vanish.
Authentication, confirmation, and witnessing narrow specific gaps; they never close them. Naming how evidence entered — commit, assert, measure, formalise — makes gaps classifiable; the taxonomy is in Four Capture Modes.
Thinking by Writing adds stream-to-summary collapse: macro-commitment (signature, final answer) without the symbolic stream that produced it.
Tier one beneath claims
Part III builds the claim ladder (tiers two–thirteen). Tier one sits beneath claims: not “very strong claims” but events reasoning is about.
A claim whose provenance never reaches tier one is assertion floating free. Tier one never is tier zero — measurements transduce; binding records are about a world that was some way regardless.
What comes next
We have substrate: fallible, distinct from the Real and from claims above — with known gaps that discipline can narrow but not erase.
Next: the four capture forces, stress test, and event schema (Four Capture Modes). Then thinking by writing — emission layers and why AI can log the stream humans usually lose (Thinking by Writing). Then the line where conjecture becomes accountable consequence (The Moment of Commitment).