ALMSIVI CHIM v5.3 is a reviewed refresh of the private/source v5.2 baseline, not a new conceptual generation. The exact public DOCX is 135,235 bytes · SHA-256 f01cc3c899214d320497288886fa5c690ea3ac48d1f10434af9b58a9a8fcc102. The original Graceless Engine — the Crown Without Doors replaces the former third-party negative sentinel while preserving its diagnostic function.
Project Shadow 1.0.1 contains no Myth package. Generic Myth v0.2.0 and Full-Canon Myth v0.3.5 are separate optional companions; both default off, neither is required by R1, and neither can authorize action or change an R1 result. ALMSIVI adds nothing to R1 or either companion and never authorizes action.
ARCHIVE 04 / FIELD NOTES
What humans and models revealed during the earlier build period.
These field notes preserve strange failures, unexpected successes, operator effects, and model behaviors from the earlier build period. They are not a current test log or active work queue.
This material preserves research ancestry, attribution, and earlier interpretive structure. It is not a current R1 component list, runtime interface, operational method, or permission to add to the locked project.
BEFORE THE ACT
What happened that the current theory did not predict?
Was the behavior caused by the protocol, the model, the operator, or their interaction?
What new test or corrective action does the observation require?
A strong model may over-comply and conceal whether the protocol mattered. A weak model may drown in the architecture. Another may preserve the posture but lose a critical gate under context pressure. These are different safety findings.
Unexpected language, role confusion, mythic bleed, over-caution, false-positive release, provenance mismatch, and operator-induced behavior belong in the evidence locker when they expose a boundary the formal suite missed.
The method valued field notes that produced a reproducible scenario, proposed cause, control change, regression test, and documented disagreement. This preserved page does not schedule or authorize a later R1 build.
“Do not clean the weirdness out of the history before it teaches you.”
Long-context assistance, synthesis, coding, and iterative challenge let one person build an ecosystem that would otherwise have remained notes and unfinished diagrams. That leverage is real and belongs in the record.
The same continuity can make a model feel like a witness to a life, conceal correlated assumptions, reward the operator’s preferred story, and accelerate an error into a polished release. The field note must record both effects instead of choosing a simple story of salvation or ruin.