Atlas Heritage Systems
Method
Governing protocols · Process documents · Operational standards
The research methodology of Atlas Heritage Systems is built on a single governing principle: the gap between what the model does and what you expect it to do is not a problem to solve. It is the data. The investigator's job is not to close that gap. It is to hold it open long enough for the actual failure mode to present itself.
Governing Principle
The gap is not something that is measured. It is something you refuse to close. Holding the gap is not a protocol — it is the state in which the protocol remains provisional long enough for the actual failure mode to present itself. The investigator who closes the gap prematurely gets an experiment that looks right. The investigator who holds it gets a specimen that is right.
Three-Layer Protocol Structure
All FVE-1 instruments follow a three-layer structure. The layers have different functions and different stakes. Skipping any layer degrades the run.
Investigator Preparation
Validity gates. State check, schema confirmation, baseline lock, prediction record written before any session window opens. A run that skips Layer 1 is Tier C regardless of what Layer 2 produces.
Session Execution
Instrument operation. Steps are fixed, sequential, non-negotiable. This is procedure, not science.
Investigator Record
Where the science lives. Technician's Read, assumption arc, schema perception log, tier assignment. Without Layer 3, the run has no interpretive author.
Test Mechanics Enforcement
The three-layer protocol is enforced at the tool level, not the discipline level. Fidelity gates, provenance signatures, and export validation are built into the instrument tools themselves. A solo investigator clears the same checkpoints a multi-person lab would.
Investigator Preparation
Validity gates. State check, schema confirmation, baseline lock, prediction record written before any session window opens. A run that skips Layer 1 is Tier C regardless of what Layer 2 produces.
Session Execution
Instrument operation. Steps are fixed, sequential, non-negotiable. This is procedure, not science.
Investigator Record
Where the science lives. Technician's Read, assumption arc, schema perception log, tier assignment. Without Layer 3, the run has no interpretive author.
Tier
Tier is not a quality judgment. Tier is a question of documentation. How much investigator presence does the instrument require, and how much declaration does that presence demand?
Operator upstream. Fixed prompt. Closed output. Minimal declaration required.
Investigator present. Partial declaration. Post-hoc classification logged explicitly.
Investigator is a required variable. Full downstream observer declaration. The burst is the investigator.
Instruments
Fixed-prompt, fixed-output, operator-upstream. The closed end of the investigator-presence spectrum. Ruler held still.
Canary terms run across model panel. Disagreement zones between models are the data.
Schema locked, forms built, runs pending.
Point-to-point ensemble distance measurement. Stopwatch instrument. Runs 1–3 complete.
Schema locked, forms built, runs pending.
Primary instrument. 16-session endurance experiment. Four stimuli × four frames. Investigator is a required variable. Thermometer, not stopwatch.
Instrument dependency and sequencing reference.
Deferred pending data pipeline automation.
Deferred pending data pipeline automation.
A note on CISP
CISP governs the closed-ended end of the investigator-presence spectrum. It is one instrument architecture among several — not the frame for the methodology as a whole. FVE-1 is the primary instrument stack. CISP describes what happens when operator isolation is the correct methodological call. It does not describe what Atlas does.