cjc0013/ouroboros-pasterski-research-program

Ouroboros Pasterski Research Program - PAUSED

A proof-carrying research campaign in which the Ouroboros AI System turned public scholarship into 49 exact, independently replayable computational result packages. The campaign ran for 18 h 11 min before the operator paused it at a stable boundary; PAUSED means the research program remains open, not that the released results are provisional.

Independent project; no affiliation or endorsement. This work is not affiliated with Sabrina Pasterski or her coauthors. They did not author, commission, supervise, review, approve, or endorse this project, and it does not represent their views. Their cited papers supplied public scholarly inputs; this repository is an Ouroboros AI System research output built from those sources.

What Ouroboros did

Ouroboros acquired and SHA-locked public source material, formed a connected mathematical work program, generated candidate extensions and corrections, rejected failed routes, retained passing results, and packaged each accepted result with executable derivations, exact checks, provenance, and post-computation assertions. All substantive discovery, derivation, testing, packaging, and manuscript production in this release were performed by the Ouroboros AI System under operator direction.

Campaign fact	Released result
Research status	PAUSED at an operator-selected stable boundary
Wall-clock interval	18 h 11 min 31 s from corpus hydration to the last completed package
Verified outputs	49 exact computational research packages
Literature coverage	34 of 48 selected papers covered; 14 remained when paused
Public inputs	36 SHA-pinned public arXiv source archives
Executable replay	49 portable derivation kernels and 51 tests
Independent validation	49/49 fresh payload matches, all falsifiers passed, and zero reference-file opens during isolated computation

Representative discoveries

• An analytic proof of a flat-space Ryu-Takayanagi coefficient equality that the source had supported numerically.
• The omitted p=4 spherical entanglement extension and its first nonzero refined-entropy term.
• An all-m transverse-nonlocality saturation theorem recognized from the m=3, m=4, and m=5 ladder and closed for every m >= 3.
• Exact corrections to the Goldilocks logarithmic pole order, the soft-Mellin factorial, and a multiparticle beta-residue factor of two.
• A self-corrected detector-sum result that records and supersedes an earlier Ouroboros false positive.
• All-dimension projective Mellin scale cancellation and projective-simplex signed-minor/rank theorems.
• Exact counterexamples and boundary theorems where broader conjectures fail.

The complete 49-row account is in RESULT_ATLAS.md, with discovery statements first and source, classification, and exact-check evidence alongside each result.

What independent replay means here

This is not a collection of copied final numbers and not a dataset of Pasterski papers. The replay starts from an empty cache, downloads the 36 exact public source archives, verifies every SHA-256, runs the 49 derivation kernels, and writes fresh symbolic or numerical payloads before loading the stored reference assertions. A separate comparison phase then checks those new outputs. Falsifiers require corrupted source bytes and mutated results to be rejected.

That procedure validates the released scientific calculations from public inputs. It does not reproduce the preceding autonomous campaign history or provide access to the Ouroboros implementation.

Where to start

1. PAPER.md and PAPER.pdf are the same paper: Markdown is the browser-readable source and PDF is the rendered form.
2. RESULT_ATLAS.md - all 49 results, claims, sources, classifications, checks, and payload hashes.
3. VALIDATION.md - what was actually tested, including the clean download replay and reference-denial audit.
4. REPLAY_STANDARD.md - the evaluator procedure and acceptance criteria.
5. references/ - machine-readable post-computation assertions; these are comparison targets, not derivation inputs.

Contents and terms

The release contains 49 portable derivation kernels, a 36-source SHA lock, 49 derivation tests plus two boundary tests, independent fetch/compute/compare/falsify commands, the paper, full result atlas, validation receipts, provenance audits, licenses, citation metadata, and a complete SHA manifest.

Human-readable attribution, licensing, access, and no-endorsement terms are consolidated in ATTRIBUTION_AND_USE.md. CITATION.cff is machine-readable citation metadata, and NOTICE is the concise legal notice preserved with redistributed code. Third-party papers are not redistributed or relicensed; replay retrieves them from their public source URLs under their original terms.