# 2026-11-01 — Stage 1: literature extraction

**Scope.** Mandate section 7, Stage 1 (structured research database) and Phase I of section 17
(literature database, research-gap report). Files written: `paper/references.bib`,
`literature/db.yaml`, `literature/records.md`, `literature/research_gaps.md`, `literature/unverified.md`.

**Method.**
1. Built a polite Crossref query helper (`query.bibliographic`, serial requests, retry on HTTP 429,
   on-disk cache) and ran four batches covering the mandate's required coverage (ECC foundations,
   ECDLP algorithms and lower bounds, signatures and side channels, SHA-2 cryptanalysis, SAT/formal
   tools, automated discovery). Each entry's title, authors, year, venue, volume/pages and DOI were
   taken from the Crossref record or the DOI landing record.
2. Items without a Crossref record were verified through IACR ePrint pages (2004/031, 2011/187,
   2012/318, 2016/382, 2020/014, 2026/232, 2024/349, 2015/350, 2014/368, 2015/143, 2017/1113,
   2008/131, 2008/270, 2014/302, 2016/374, 2009/541), the arXiv API (2406.20072, 2408.06292,
   2506.13131), official standard URLs (SEC 1 v2.0 and SEC 2 v2.0 title pages, FIPS 180-4 PDF,
   FIPS 186-5 DOI, RFC 6979, BIP 340/341 headers from the bitcoin/bips repository, bitcoin.pdf,
   SafeCurves, libsecp256k1 `src/scalar_impl.h`), the CEUR-WS Vol-3717 index, the JKU record of the
   SAT Competition 2020 description, the Certicom ECC Challenge PDF (solved-challenge tables) and
   web searches for the Satoh–Araki 1998 bibliographic data.
3. Numeric claims used in `db.yaml` / `records.md` (step counts, complexities, hardware, dates) were
   read from abstracts (ePrint/arXiv) or from the Certicom tables, not from memory. Where a requested
   figure could not be found (the "117.35-bit" ECDLP, a post-2009 prime-field record, Blackburn–Murphy
   1998) it was left out and documented in `unverified.md`.
4. A scratch LaTeX build (`\nocite{*}`, `aer.bst`, pdflatex + bibtex + pdflatex ×2 in /tmp) was run
   to make sure the bibliography compiles without BibTeX errors or warnings; see the build note below.

**Outcome.** 129 BibTeX entries, all verified (see the summary in the final report); 82 database
items; one ECDLP-records table and one SHA-2 records table; a two-page gap report that lists which
mandate claims are already settled (√6 endomorphism bound, no sub-exponential prime-field ECDLP,
Schnorr/ECDSA proof structure, SHA-256 frontier 31/37/39 steps) and eight candidate directions D1–D8
for Stage 2.

**Corrections to the brief discovered during verification.** LadderLeak is by Aranha, Novaes,
Takahashi, Tibouchi and Yarom (key kept as `aldaya2020ladderleak` at the coordinator's request);
Lee–Cheon–Hong 2011 is ePrint 2011/187 with a different title; Cheon–Hong–Kim's J. Cryptology 2012
paper is titled "Accelerating Pollard's rho algorithm on finite fields"; the 2016 FPGA record is
sect113r2 with no "117.35-bit" figure; "Bailey et al. 2009" describes a planned attack on ECC2K-130
that the Certicom document still lists as unsolved.

**Next.** Stage 2 hypothesis generation from `research_gaps.md` section 4; the discovery log
(`research_log/discovery_log.jsonl`) starts when the first hypothesis is evaluated.

**Build note.** `/tmp/bibtest`: `pdflatex` + `bibtex` (BibTeX 0.99e, TeX Live 2026, `aer.bst`) +
`pdflatex` ×2 with `\nocite{*}` produced 129 `\harvarditem` entries, BibTeX exit 0, no warnings, no
undefined citations. Verification tally: 109 entries carry a DOI, 18 are URL-verified
(ePrint/arXiv/standards/BIPs/software), 2 have neither (Satoh–Araki 1998, confirmed through citing
sources; the SAT Competition 2020 description, confirmed through the JKU publication record).