JAIC SUBMISSION NOTES Declaration and Persistent State Change in Artificial Systems Prepared for Joseph Anady FILES AND THE CURRENT UPLOAD SCREEN JAIC_Manuscript_Joseph_Anady_Revised.docx is the complete editable main manuscript. On the Attach Files screen shown, use Browse to select this Word file, wait for the upload to complete, and then choose Proceed. JAIC_Manuscript_Joseph_Anady_Revised.pdf is the matching reference PDF rendered from that Word file. Do not upload the Word file and PDF as duplicate main manuscripts unless Editorial Manager specifically requests both. JAIC_Supplementary_Material_S1_Revised.zip is the reproducibility supplement. Attach it under the supplementary material item type when that option is offered. It is not the single main manuscript requested on the initial upload screen. Do not approve the final submission without attaching S1, because the manuscript refers to it. This notes file is for the author and is not part of the manuscript. It should not be uploaded as the main article. PORTAL METADATA Title Declaration and Persistent State Change in Artificial Systems Author Joseph Anady Affiliation Independent Researcher, Missouri, United States Correspondence [author/contact email redacted] Abstract (135 words) I examine the measurement problem in an account of consciousness that places declaration before an internally constructed change. I analyze 167 public adapter checkpoints from one training run, with 792 gradient observations and 79 evaluation losses. Comparing the effective weight update with its factors separates operator change from equivalent factorizations. The first saved nonzero update occurs at step 2, the gradient maximum at 169, and a descriptive split after 181. At a factor amplitude threshold of 0.002, all three lags select step 190 for both the output factor and the effective update; without filtering, two select 725. Two principal components explain 94.990% of sampled output factor variation. These results establish persistent parameter change, not a measured declaration or subjective experience. I specify independent event detectors and a prospective protocol separating temporal prediction from causal feedback. Keywords (6) machine consciousness; self reference; declaration; persistent state change; LoRA; reproducibility Running head Declaration and Persistent State Change Section to select when available Assessing AI Consciousness The journal's published scope includes this named area. The paper is framed as a formal hypothesis and a methodological analysis of measurement, not as an empirical demonstration that an AI is conscious. Use the closest actual option displayed by the portal if its labels differ. COMPLETED REVISION The manuscript preserves the original sequence proposed by the author: declaration as a marker, a subsequent internally constructed change proposed as relevant to consciousness, and continued processing from the changed state as cognition. The original completed numerical analysis remains separate from that proposed interpretation. The introduction states the hypothesis directly. The formal section defines the system state, center, recurrent process, inputs, discrepancy, and independent event detectors. The timing definition can report change before declaration, declaration before change, or simultaneity at the measurement resolution. A conditional search only after declaration is explicitly distinguished from an onset ordering test. The theoretical comparison discusses self models, attention schema theory, higher order accounts, global workspace models, and the computational indicator approach. The foundational emergent misalignment paper and its published extension are cited. The approximately 95 percent comparison is attributed specifically to Figure 8 in Section 4.1 of Turner et al. (2025, version 1). The body uses more readable numerical precision. Exact numerical anchors, source identifiers, and executable commands remain available. Repetitive disclaimers have been reduced without changing the limits of the evidence. Punctuation There are no em or en dashes in the manuscript. Prose has been recast rather than relying on dash punctuation. Mathematical minus signs remain, as do necessary hyphens in exact repository names, commands, author surnames, and published reference titles. Removing those would damage equations, reproducibility, or attribution. CHECKED DELIVERABLES The matching PDF has 24 pages. The Word file contains a 135 word abstract, six keywords, 18 numbered display equations as editable mathematics, four embedded figures, three data tables, and 17 reference entries. The figure images are 3000 pixels wide and displayed at five inches, giving 600 dpi. Corresponding EPS files are included in S1. Every final rendered page was visually reviewed. The final PDF was checked for replacement glyphs and text outside the page. The Word package was checked for invalid archive entries, missing equation labels, tracked changes, comments, and broken exact source identifiers. All three offline commands in Appendix A were executed successfully on the delivered supplement. All five transferred data artifact digests match the original receipts, and all 22 computational tests pass. The supplement includes the new execution logs and the current environment record. Its delivery checksum manifest and ZIP integrity were verified. The original data, analysis programs, source receipts, notebook, and figure files are retained. These are computational and file integrity checks performed with AI assistance. They are not independent human replication. Raw source acquisition was not repeated, no new declaration was measured, and the prospective intervention protocol was not executed. The original notebook is preserved rather than presented as newly executed. JOURNAL REQUIREMENTS CHECKED The published submission instructions require online submission through Editorial Manager, a title and an abstract of no more than 150 words on the front page, and no more than six keywords. Word and TeX preparation are preferred. The figure guidance prefers a width of at least 3.5 inches, readable Times Roman lettering of at least 9 points, and EPS at 300 dpi or higher. The manuscript uses the current 150 word limit rather than a more permissive older template example. The current web submission instructions, Word template information page, aims and scope, and AI authorship policy were inspected. The linked downloadable Word template archive could not be freshly retrieved in this revision. Therefore the existing journal layout, page geometry, and author date reference structure were preserved, not represented as independently checked against every file in that archive. The AI disclosure identifies ChatGPT and describes its substantial use for source inspection, code, numerical execution and interpretation, figures, mathematical exposition, drafting, and revision. No AI is listed as an author. The disclosure does not claim that the author has already completed final approval. Official sources checked for this revision https://www.worldscientific.com/page/jaic/submission-guidelines https://www.worldscientific.com/page/jaic/stylefiles-readme-doc https://www.worldscientific.com/page/jaic/aims-scope https://www.worldscientific.com/authors/position-statements/ai-tools AUTHOR CONFIRMATIONS BEFORE APPROVAL Confirm that the author name, affiliation, and correspondence email are correct. Supply an ORCID only if it is yours. No ORCID or institutional appointment has been invented. Answer the funding and competing interest questions from your actual circumstances. No assertion of no funding or no competing interests has been invented. If the portal requires those declarations within the manuscript rather than only in its fields, add the confirmed statements before approval. Confirm that the manuscript is not under consideration by another journal or publication. This is an explicit condition in the journal's submission guidance and cannot be inferred from the files. Read the revised hypothesis and event definitions and approve them as your intended scientific claims. Review the AI disclosure and accept responsibility for the final submitted material. Personally running the Appendix A commands provides an additional author check; this package does not claim that you have already done so. Before choosing final Approve, open the PDF generated by Editorial Manager itself. Check the title, author details, all equations, four figures, three tables, references, and the presence of S1. The attached reference PDF does not establish that a different server rendering will be identical. No manuscript preparation can guarantee acceptance, a particular review outcome, or that an editor will never request a formatting change. The files address the verified public instructions and identified manuscript issues; scientific judgment and additional portal requirements remain with the journal.