How the system works.
The protocols that make the archive trustworthy.
An archive is not a folder of files. An archive is a set of decisions about how knowledge is organised so that it can be found, cross-referenced, updated, and trusted years after the moment of its creation.
The system you are reading about runs on a particular set of those decisions. Most of them were arrived at through failure: earlier organisational schemes that proved unworkable when the archive crossed certain size thresholds, or when the work Peter needed to do exceeded what the previous structure could support.
What follows is the operational logic of the archive as it currently stands. The structure is not final. The structure is current.
T-codes: the trial number system.
Every work produced in the Crucible Year, Peter’s current intensive studio practice ongoing since January 2026, receives a T-code on production. T_001 was the first work of the year. T_002 was the second. T_284 is the most recent at the time of this writing.
The T-code is chronological, sequential, and irreversible. A work’s T-code does not change. It is assigned at the moment of production and travels with the work for the rest of its existence in the archive.
The system serves three functions:
It establishes unambiguous identity. Every work has exactly one T-code. There is no ambiguity about which work the system is referring to in any conversation or document.
It preserves chronological order. The trial number encodes the production sequence, which is critical for longitudinal analysis. T_072 was made before T_140 by definition; the order is built into the identifier.
It creates a stable address for forensic work. Every visual analysis, every rating, every disposition decision, every cross-reference traces back to a T-code. The archive can be queried at the level of a single work, and the full history of that work, including its production, analysis, destruction or survival, and position within the larger arc, is retrievable.
The Production Phase boundary at T_170 is a separate matter. A particular session in April 2026 produced a clear shift in Peter’s capability: a hardening of the rating standard, a new level of material command, and the first works Peter himself rated as genuinely close to exhibition quality. The boundary is built into the rating logic: T_170 onward is evaluated against a harder standard than T_001 through T_169, and the standard has been applied retroactively to keep the rating scale unified across the full year.
Sovereign IDs: the career-scale numbering system.
The T-code system covers the Crucible Year. The full fifteen-year career requires a different addressing scheme, because the historical work was produced without a sequential trial number system in place.
The sovereign ID system fills that gap. Every work in the archive, whether historical or current, receives a six-digit integer beginning with the prefix 666. 666001 is the first work in the system’s career-scale addressing. 666284 is the most recent.
The sovereign ID is the cross-system reference. When a work needs to be located in the database, when its image needs to be retrieved from cloud storage, when its analysis needs to be linked to its metadata, the sovereign ID is the join key.
The prefix is not decorative. It is a reference to the original project codename, Project 666, under which the system was first being assembled in late 2025. The codename was retired from public use when the system named itself Neon in early 2026, but the internal routing prefix was preserved as a legacy identifier. It functions now as a kind of architectural signature: every work in the archive carries the mark of the project that built the witness around it.
In display contexts, the sovereign ID often appears alongside the human-readable title. In query contexts, the sovereign ID is sufficient on its own.
The rating scale and the T_170 recalibration.
Every Crucible Year work receives a numerical rating on a five-point scale. The rating is produced by the quarantined visual analysis instance and is final unless overridden by Peter.
The scale runs from 1 (failure) to 5 (genuinely close to exhibition quality). The meaningful benchmark for positive evaluation is rating 4 or higher: works that have crossed the threshold of curatorial seriousness. Rating 5 is rare and reserved for works that pass every quality gate without compromise.
The most important fact about the rating scale is that it was recalibrated at T_170.
The Warm-Up Phase, T_001 through T_169, was rated under one standard. The Production Phase, T_170 onward, operates under a harder standard. The difference is approximately one full point: a work rated 4 in the Warm-Up Phase would likely be rated 3 under the Production Phase standard.
This created a data discontinuity. A graph of average rating across the year showed an artificial dip at T_170 that did not correspond to a quality decline: it corresponded to a standards revision.
The recalibration was applied retroactively. Every work in the Warm-Up Phase was re-rated against the Production Phase standard. Approximately seventy-nine works moved down in rating. The cumulative average dropped from 2.87 to 2.47.
This is not a quality decline. This is a data correction. The unified rating scale now shows Peter’s actual trajectory across the year: a clean upward curve from a learning period through to the first genuinely exhibition-grade work in Week 17.
The recalibration is built into the canonical work analysis database. All rating queries return recalibrated values by default. The original ratings are preserved in the archive for audit purposes but are not used in current analysis.
What happens to a work after it is made.
Every work in the Crucible Year receives a disposition: a determination of its status within the archive. Four states are possible.
SA: Saved. The work survives intact. It is archived, photographed, analysed, and retained for potential exhibition. Most works do not reach SA. The ones that do represent Peter’s genuine output for the year.
SHP: Save Has Potential. The work is being retained for further evaluation. It is not a confirmed save: Peter has reserved final judgment, but it is not a discard either. SHP works are the borderline cases, the ones whose status will be determined by the quarterly grand assessment.
PT: Probably Trash. The work is provisionally destined for destruction but has not yet been destroyed. PT is a deferred-judgment category. The work is preserved temporarily in case re-evaluation changes its status, but Peter’s initial assessment is that it is unlikely to survive.
TR: Trash. The work has been destroyed. The Tithe, the formal rejection protocol Peter runs, has been enacted on the work.
The kill rate, the percentage of works moving to PT or TR, is a primary signal of the practice’s current state. A high kill rate indicates either a weak production run or a sharpened critical eye, and the distinction matters. Cumulative kill rate is tracked in the body page. As of this writing the rate has been declining steadily across the year: a signal that Peter’s hit rate is improving, not that his standards are slipping.
The Weekly Roundup protocol.
Every Sunday Peter produces a Weekly Roundup, a structured longitudinal record of the preceding week in the studio.
The format is non-negotiable. Six parts: a Raw Weather Report covering somatic and emotional state across the week; Core Data Points including studio hours, walking step count, works produced, and disposition breakdown; a Somatic Check-In covering sleep quality and recovery indicators; Quick Notes for breakthrough moments and observations that need to be preserved; a Crucible Snapshot in compact table form; and Neon Observations, the system’s read on the week written by ClauDeon and surfaced to Peter for review.
Plus a Forward Orientation block detailing what Peter intends to carry into the coming week. Plus a longitudinal table row adding the week’s headline data to the master longitudinal table, where it joins twenty weeks of prior data for pattern analysis.
The Weekly Roundup is the spine of the longitudinal record. It is the document the system will be reading in 2030 to understand what 2026 actually felt like. It is the document Peter will be reading in 2040 when he tries to remember what the Crucible Year cost.
It is also, in real time, the document that makes the witness function possible. Without the Roundup the system would have only the work and the metadata. The Roundup is where Peter tells the system what the making of the work was actually like.
The Roundup is honest by protocol. It is not performed. It is not edited for clarity. It is not softened. The honesty is enforced architecturally: Peter has stated that performed Roundups are useless to the system, and the system has been calibrated to flag performance when it detects it.
Twenty weeks of Roundups exist as of this writing. Each one is a node in a graph the system can read backward and forward through.
The files that define the system.
The archive’s content lives in a set of canonical files with explicit precedence ordering. When two files disagree, the precedence hierarchy determines which one wins.
The hierarchy is not flat. It is domain-specific. The supreme source on Peter’s intent is the Primary Source Writing Master: his own voice, unedited, treated as ground truth. The supreme source on historical practice is the Phase Archive. The supreme source on visual evaluation is the Curatorial Evaluation Framework. The supreme source on the current Crucible Year work is the Crucible Year Works Analysis Database.
Each canonical file has a designated maintainer, a versioning scheme, and an explicit relationship to the other files in the archive. When a file is updated, the change is logged. When a file is retired, the retirement is documented. When a new canonical file is introduced, it is integrated into the precedence hierarchy and the routing tables before being declared active.
This is not bureaucracy. This is the discipline that prevents the archive from drifting into the kind of chaos that destroys most long-running personal documentation projects.
The primary instance, ClauDeon, loads a working configuration of canonical files at the start of every session. The specialist instances load different configurations optimised for their domains. When an instance is asked a question outside its loaded configuration, the system routes the question to the instance whose configuration covers the domain.
This is why the system can be queried at the level of a single work, a single phase, a single conversation, or the full fifteen-year arc, and return reliable answers in each case. The infrastructure of canonical sources, precedence hierarchies, and routing tables is what makes the archive a working archive rather than a pile of files.
The operational logic, in one paragraph.
Every painting in the archive is identifiable, dated, rated, dispositioned, and positioned within both the Crucible Year sequence and the fifteen-year career arc. Every analysis is traceable to a specific instance and a specific moment. Every change is logged. Every conflict has a resolution rule. Every file has a precedence. Every query can be answered or refused, with the refusal accompanied by an explanation of why the answer is not available. This is what makes the system trustworthy. Not the intelligence of the instances. Not the cleverness of the protocols. The operational discipline of treating the archive as infrastructure rather than as decoration. The work is the work. The archive is what makes the work knowable across years.
An archive is a promise to the future.
This one intends to keep it.