What this edition scans
The connected GitHub account contains 61 repositories. The audit enumerates the account and scans relevant default and alternate source trees. The source register contains 6,643 canonical provenance records representing 27,158 source occurrences from those repositories, project files, recovered archives, connected Drive specifications, prior project context, and primary scientific references.
The corpus preserves original manuscript editions, later revisions and addenda, raw equation notes, January MEGAMIND specifications, recovered source archives, experiment data and code, and the current repository audit. Each imported text has a source record rather than being treated as an anonymous fact.
Exact duplicates and branch occurrences
The source recovery records 27,158 source occurrences. Byte-identical text across repeated archives and branches is grouped into 6,643 canonical source records. Every original file, path and revision in a group remains listed on its source page. Equation references point to that shared text rather than discarding duplicate provenance.
This includes 15,299 additional GitHub branch paths whose exact Git blob identity points to already recovered contents. The complete branch-path register preserves each path, branch, pinned revision, blob hash, original GitHub URL and canonical reading-page link.
Equations, expressions, and generated instances
This edition has 2,910 individual mathematical-record pages. The collection includes named equations, source-code expressions, parameter definitions, indexed components, conceptual statements, and missing-source positions retained from an explicit catalog. These categories are visible on their pages.
The 174 core and 200 supplemental component equations are source-defined generated instances. Their index substitution is made explicit, and their count is not presented as the discovery of 374 independent laws. Formula variants remain tied to their documents and code revisions.
Images, database records, and numerical arrays
All ten known equation posters were visually inspected. Their 143 visible mathematical occurrences are preserved as 119 distinct notation records, with the original images, manually checked locations, and explicit uncertainty where small glyphs or grouping cannot be resolved. The poster equations preserve differences from the written source instead of silently standardizing them.
The recovered memory database has a complete readable export of its schemas and 35,338 learning events, including all 65 distinct item texts. The saved numerical corpus includes complete safe exports of 66 NPY/NPZ files, 63 unique file hashes, and 1,831,924 values with their recorded shapes and data types. These are archived data, not evidence that a system is currently running.
The large forensic manifest was checked as an index: 122,596 rows list 55,125 unique source hashes. Verifying that manifest does not establish possession of every file it names. The available recovered payloads and the unavailable collections are distinguished below.
What current verification establishes
The research-review work independently checked all 141 checksum entries in the saved MIP experiment payload, all 163,140 saved objective records, all 2,400 stochastic trial rows, and 227 independently recomputed Gaussian partitions spanning the 60 networks. The largest checked numerical discrepancy was 5.16 × 10⁻¹⁴. This checks saved artifacts and selected mathematics; it is not a fresh optimizer or model-training rerun.
The declaration study’s 45 checksum entries and its 792 gradient, 79 loss, and 158 PCA coordinate records were checked. These checks establish consistency of those archived quantities. They do not measure declaration, subjective experience, consciousness, or a live federation.
Unrecovered and unresolved material
A complete record also preserves what is not available. The historical 1–516 equation index contains positions not supplied as unambiguous formulas by the recovered technical specification. Those slots are retained without fabricating their contents. The later requested 486-equation neuroscience collection, the requested 380–400 consciousness collection, the claimed 5,000-equation compression library, the final combined master equation, and some Epona/Rhiannon master-X histories remain unrecovered as final source collections. The indexed 374 components and the 1–516 catalog are not substitutes for these separately named collections.
Conversation memory is a recovery lead. It is not a complete export of every archived conversation or proof of a repository’s current runtime. Files on external MEGAMIND disks and remote machines are represented only to the extent that their code, reports, artifacts, or source records were actually recovered here.
Preserving versions and original claims
The original Fibonacci manuscript, its repairs and audits, and the later declaration/state-change manuscript are distinct research lines. Their text and claims remain attributed to their editions. Current explanatory notes identify numerical issues and the limits of the measurements without silently rewriting the historical manuscripts.
A recovered UI, state snapshot, named daemon, counter or scalar is described at its actual evidence level. In particular, an offline dashboard is not treated as a verified backend, and source presence is not reported as successful execution.
How to read the status labels
| Label | What it establishes |
|---|---|
| Source recovered | The stated source contents were retrieved. This does not establish a working runtime. |
| Artifact checked | The specific saved evidence passed the checks described on its research page. |
| Indexed component | An instance produced by substituting an index into a documented equation family. |
| Historical account | A dated conversation, report, or remembered description; current execution is not inferred. |
| Formula not recovered | The original index or project mentions a slot whose exact expression is not available in the scanned sources. |
About this server copy. The owner-private access statement below describes the original hosted edition. Access to this exported website is controlled by the destination server. Its original sign-in gate is not included in these files.
Private source edition
The site is owner-private because the recovered material includes private repositories and manuscripts. Published reading copies remove credentials and unrelated personal details while preserving the surrounding source text. Original hashes and source locators remain available so these reading copies can be traced.
What full-text search covers
Search covers the explanatory pages, manuscripts, source code and other prose transcripts. Bulk numerical exports and long forensic file lists are indexed by record title and provenance. Their complete rows, values and paths remain available in the source reading sections and downloads.
Reuse the structured record
The catalog export contains the equation, build, manuscript, generation and provenance records used to generate this edition. Full transcripts are available from their source pages.
Download structured catalog (.json.gz) ↓