The Input Stage Before Curiosity: Fragment Retention, Source Multiplicity, and the Cost of Angle (v0.1)

The Input Stage Before Curiosity: Fragment Retention, Source Multiplicity, and the Cost of Angle (v0.1) VERSION: 0.1 PUBLICATION DATE: 2026-10-21DOI (this version): 10.5281/zenodo.23536029 (none — first deposit, its own record)UPLOAD TYPE: publication / workingpaper LICENSE: cc-by-4.0 RELATED IDENTIFIERS: cites 10.5281/zenodo.23366898 cites 10.5281/zenodo.23535380 cites 10.5281/zenodo.23394727 cites 10.5281/zenodo.23368185 cites 10.5281/zenodo.23534995 cites 10.5281/zenodo.21819489 CREATOR (роль EDITOR — стоящая договорённость): Mishko, Nikolai | Astana Digital Hub, Kazakhstan A short structural paper on what stands BEFORE the first arrow of the corpus's canonical account of where new goals come from. W37 section 1.1a states that chain as external-source surprise (W43) → question horizon → question-generation rate → goal candidates → goal acceptance → endogenous goal expansion. This document states one claim about its entrance, names two repairs that follow, and performs neither: the repairs belong to W43, which owns the gate. IT PROPOSES NO MEASURE AND STATES NO THEOREM. Every structural claim is cited to its owner. No weight is placed on any source and no count of sources is given; any number here would price a novelty source, which the corpus has declined twice over — as underived pseudo-accuracy and as anti-capture — across five review rounds. 1. THE GATE IS DEFINED PER EVENT, SO SUB-THRESHOLD MATERIAL NEVER REACHES IT. W43 classifies an event as a surprise when the prediction error exceeds a threshold AND the source is certified external, with the error measured as surprisal in bits under the RECEIVING AGENT'S OWN predictive model. Below that threshold nothing is provided for. But a fragment that is unremarkable on arrival can become load-bearing later, when it is matched against a fragment held from another domain: taken alone its error is below threshold, taken as a pair the two carry a discrepancy neither carried alone. MATCHING IS A RELATION BETWEEN RETAINED FRAGMENTS; A THRESHOLD IS A PROPERTY OF ONE EVENT. The gate therefore drops, by construction rather than oversight, exactly the material on which cross-domain matching operates. Neither W43 nor W37 contains retention of sub-threshold material or any recombination operator; this was checked against both rather than inferred. Lowering the threshold does not help — it admits more single events and still admits no relation between them. What is absent is a stage, not a setting. 2. THE EXTERNALITY CERTIFICATE IS BINARY AND DOES NOT COUNT SOURCES. Ext equals one exactly when write authority is absent and provenance is external. The certificate is per source and boolean, so ONE external source and TWENTY DIFFERENTLY POSITIONED EXTERNAL SOURCES YIELD THE SAME VALUE: cardinality and position are not registered. The package's own manifest already warns that the certificate establishes externality and not statistical independence of the source's residual information; the present point is adjacent and distinct. W41's Theorem 1 shows that an independent element requires a source outside the agent's reachability horizon, and one such source satisfies that — but it does not follow that one source outside the reach is as good as many differently positioned ones, and nothing in the chain can state the difference. 3. CONTRADICTION BETWEEN ANGLES IS POSITION-RELATIVITY, NOT A DEFECT OF THE SOURCES. More sources give a fuller picture and a more contradictory one. In the corpus's own terms this is already implicit and its consequence is not drawn: the discrepancy is surprisal under the receiving agent's own model, so it is position-relative by definition. Two differently positioned receivers assign different values to one event and the threshold can fall between them — THE SAME EVENT IS A SURPRISE FOR ONE AND NOT AN EVENT AT ALL FOR THE OTHER. The chain has no form for this, for a reason it states itself: W37's fourth consequence bounds the whole architecture to a single agent, and the obstruction to an ensemble form is proved rather than open. So contradiction between angles is not noise to be reduced: a reconciliation step that removes it removes the thing that was gained by having several sources. 4. A CANDIDATE, STATED AT THAT STRENGTH DELIBERATELY. Matching a fragment from one domain against a fragment from another changes which problems the agent treats as its own — a movement of scope, not a metaphor for one. By W33's no-aggregation theorem the ensemble goal-generation rate turns on whether scopes move, that information is absent from per-agent rates at any resolution, and the quantity must be measured on the collective goal process as an object in its own right. No such measurement exists. The intake stage described here is therefore a CANDIDATE agent-level mechanism of scope movement. It is not a measurement of scope movement, it does not close the open question, and it does not supply the ensemble form that is proved unreachable from per-agent instrumentation. What it offers is a place to look on the single-agent side for the thing the ensemble side cannot assemble. TWO REPAIRS ARE NAMED AND NEITHER IS PERFORMED, both owed to W43: whether the classification should admit a retained SET as a bearer of discrepancy alongside the single event it admits now, and what externality means for a set whose members have different sources; and whether the externality certificate should carry source structure — at minimum cardinality, and whether two certified sources occupy distinguishable positions — as a companion to the boolean it carries now. Neither asks for a weight or a sufficiency count: the question is whether structure is REGISTERED, not how much of it is good. A NOTE ON THE RECORD. This document stands in a record of its own and is not a version of the Surprise–Novelty package, which keeps its own line. The cost is named inside: a reader reaching that package does not arrive here, and the repairs named here are owed by a document that lives elsewhere, so the two can drift apart by version. Each repair accordingly names its owner, its section, and the version of W43 it was read against. STATUS: first issue, NO REVIEW ROUND. Stated rather than smoothed over. The W-number is unassigned; the corpus W-line stops at W44.---------------------------------------------------------------------- NOTES FIELD:First issue, no review round, and that is stated in the document rather than smoothed over. The author's role on this record is EDITOR: the text is drafted by AI systems, reviewed by several other AI systems, and edited by a human who adjudicates disputed points. Zenodo's creator model cannot express that arrangement, and Editor is the most honest of the available labels. https://zenodo.org/records/18859746 Структура Symbiotic Codes: полный пакет документации (SCF) 10 документовhttps://zenodo.org/records/18848521 Symbiotic Codes Framework: Расширенный пакет документации (SCF) 4 документаhttps://zenodo.org/records/20176518 Рабочий доклад Демографический коллапс городовhttps://zenodo.org/records/20175670 Рабочий доклад Кризис избыточности человеческого потенциала (HRC)https://zenodo.org/records/20175186 Рабочий доклад Ловушка зависимостейhttps://zenodo.org/records/20174938 Рабочий доклад Ловушка ресурсовhttps://zenodo.org/records/19352775 Рабочий доклад v7.0 Цивилизационные ловушки: Единая динамическая теория стабильности, краха и симбиотического переходаhttps://zenodo.org/records/20157347 W11 v2.0 — Типология экологических кризисов и стабильность цивилизации: доказательство Ляпунова, спектральное объединение MFLS, стохастическая динамика и физический субстратный слой — математическое расширение с выравниванием W20 v8.3, упорядочиванием приоритетов Kodex и триадной архитектурой управленияhttps://zenodo.org/records/20153836 W12 v2.0 — Многоуровневая топология сети AGI и стабильность цивилизации: триадная блокировка, когнитивное разнообразие, спектральный критерий MFLS и незаменимость человеческой новизны — формальная теория с выравниванием W20 v8.3, упорядочиванием приоритетов Kodex и топологией бассейна SCFhttps://zenodo.org/records/23464603 W13 Открытый доступ Камеральная калибровка инженерных априорных оценок: методология и предварительное заключение — калибровочные реперы W13_BRIDGE, аргументы переноса, проектные константы, карта поиска, досье доказательств и итоговое заключение CR_scihttps://zenodo.org/records/23465315 W14 v2.0 — Динамика и управление цивилизацией: Единая теория управления, теория игр, бифуркация и замыкание MFLS (согласовано с трехкорпусной архитектурой W20 v8.3)https://zenodo.org/records/20179364 W20 v8.3 — Omega-u Civilizational Framework: Unified Multi-Trap Synthesis, Triadic Control Architecture, Cross-Cascade Matrix, and Symbiotic AGI Governance — Master Theorem with Full W11–W19 Alignment, Bootstrap Protocol v1.2, and Ten AGI_FROZEN Pathshttps://zenodo.org/records/20151370 W19 v2.4 — Ловушка информационной асимметрии: эпистемический коллапс, когнитивный захват и дефицит сигналов в AGI — Формальная динамическая модель с выравниванием W20 v8.3, упорядочиванием приоритетов Kodex и триадической архитектурой управленияhttps://zenodo.org/records/20151790 W18 v2.2 — Ловушка технологической стагнации: убывающая отдача, захват инноваций элитой и вычислительный парадокс — Формальная динамическая модель с выравниванием W20 v8.3, упорядочиванием приоритетов Kodex и топологией бассейна SCFhttps://zenodo.org/records/20151979 W17 v2.4 — Ловушка институциональной деградации: динамика распада Бенкво, коррупционная ловушка и институциональное ограничение перехода AGI — Формальная динамическая модель с выравниванием W20 v8.3, упорядочиванием приоритетов Kodex и топологией бассейна SCFhttps://zenodo.org/records/20152256 W16 v2.2 — Ловушка финансовой нестабильности: автокатализ долга, циклы заемного финансирования, ловушка Фукидида как равновесие Нэша и риск вымирания цивилизации — Формальн