Historical dossiers
116
Woeginger register, 1986–2016
The shape of the literature
Register statistics
The historical register and the post-2016 census show how claims, attention, and the wider literature changed over time. Every chart keeps its source receipt and exact data beside it.
The numbers first
Post-2016 candidates
43
Censused, awaiting review
Never examined
86.2%
Historical claims with no recorded adjudication
Indexed works, 2020s
2,741
2021-2026.5, against 369 in the whole prior decade
Dead links
45 of 174
Printed links now returning errors
What the data says
- The literature outlived its referees. Journal output has stayed in the tens of works per decade since 2001, while 2,741 indexed works landed in the partial 2020s against 369 in the whole prior decade. New claims now live on preprint servers first, refereed later, if ever.
- Attention comes in waves, and the wave broke in 2016. The busiest stretch was 2003 to 2012, peaking at 14 claims in 2011. The quiet after 2016 is real, and it is exactly when the field’s only scorekeeper stopped.
- Hardly anyone ever looked. 100 of the 116 historical claims, 86.2%, carry no recorded verdict. That is a fact about review capacity, not about the claims.
- The claimants changed their minds. P = NP ran at 52.6% of historical claims; since 2016, 55.8% of candidates claim the opposite, drifting toward the position about 80% of polled experts already held in 2012.
- The record is rotting. A quarter of the printed links, 45 of 174, are already dead. Preservation is not a courtesy; it is most of the job.
The refereed heartbeat and the preprint explosion
Provenance: OpenAlex API, retrieved 2026-07-25, mailto-identified; strict phrase title_and_abstract.search; journal = primary_location.source.type:journal + type:articleJournal-published works
Independent scale: 0–90 works
All indexed works
Independent scale: 0–2,800 works
What this means: the refereed heartbeat never sped up; the explosion is preprints. A register that watched only the journals would miss almost everything.
* 2021–2026.5: partial decade (5.6y); all_works dominated by 1477 preprints; OpenAlex preprint indexing expanded over period - multiplier soft, direction real.
| Decade | Journal-published | All works | Note |
|---|---|---|---|
| 1971-1980 | 2 | 2 | Full decade |
| 1981-1990 | 5 | 8 | Full decade |
| 1991-2000 | 9 | 17 | Full decade |
| 2001-2010 | 42 | 129 | Full decade |
| 2011-2020 | 82 | 369 | Full decade |
| 2021-2026.5 | 49 | 2,741 | partial decade (5.6y); all_works dominated by 1477 preprints; OpenAlex preprint indexing expanded over period - multiplier soft, direction real |
Forty years of claims
Provenance: Woeginger register (frozen 2016) + post-2016 census v2 (discovery frame in sweep/DISCOVERY-FRAME.md)What this means: attention comes in waves. The 2003–2012 crest was the busiest stretch this problem has ever had; the shallow water after 2016 is when the register froze. A quiet register never meant a quiet problem.
Show the exact yearly counts
| Year | Claims |
|---|---|
| 1986 | 1 |
| 1987 | 0 |
| 1988 | 0 |
| 1989 | 0 |
| 1990 | 0 |
| 1991 | 0 |
| 1992 | 0 |
| 1993 | 0 |
| 1994 | 0 |
| 1995 | 0 |
| 1996 | 1 |
| 1997 | 1 |
| 1998 | 0 |
| 1999 | 0 |
| 2000 | 1 |
| 2001 | 1 |
| 2002 | 1 |
| 2003 | 5 |
| 2004 | 7 |
| 2005 | 11 |
| 2006 | 9 |
| 2007 | 3 |
| 2008 | 9 |
| 2009 | 9 |
| 2010 | 11 |
| 2011 | 14 |
| 2012 | 12 |
| 2013 | 4 |
| 2014 | 6 |
| 2015 | 4 |
| 2016 | 6 |
| 2017 | 2 |
| 2018 | 8 |
| 2019 | 2 |
| 2020 | 8 |
| 2021 | 4 |
| 2022 | 4 |
| 2023 | 3 |
| 2024 | 4 |
| 2025 | 7 |
| 2026 | 1 |
The direction flip
Provenance: Woeginger register (frozen 2016) + post-2016 census v2 (discovery frame in sweep/DISCOVERY-FRAME.md)- P ≠ NP
- P = NP
- Other
| Era | P ≠ NP | P = NP | Other | Total |
|---|---|---|---|---|
| Woeginger era | 49 (42.2%) | 61 (52.6%) | 6 (5.2%) | 116 |
| Post-2016 | 24 (55.8%) | 17 (39.5%) | 2 (4.7%) | 43 |
Who ever looked?
Provenance: Woeginger register (frozen 2016) + post-2016 census v2 (discovery frame in sweep/DISCOVERY-FRAME.md)- Adjudicated or refuted
- Withdrawn
- Unadjudicated
What this means: most claims were never checked by anyone, in either era. The post-2016 numbers look better partly because those claims are young: attention fades as claims age. The eras are not perfectly comparable either, because Woeginger recorded verdicts only inside his commentary, so the historical numbers come from reading his notes.
| Era | Adjudicated or refuted | Withdrawn | Unadjudicated | Total |
|---|---|---|---|---|
| Woeginger era | 16 (13.8%) | Not separately recorded | 100 (86.2%) | 116 |
| Post-2016 | 12 (27.9%) | 4 (9.3%) | 27 (62.8%) | 43 |
One quarter of printed links are dead
Link rot
45 of 174
Printed links returning an explicit 4xx response or another held dead-link result.
What this adds up to
Attention, not refutation
86.2% of historical claims carry no recorded adjudication. The register’s biggest fact is about the field’s attention, not the claims’ merit.
The direction flipped
Claimants ran 52.6% P = NP in the Woeginger era and 55.8% P ≠ NP after 2016 — the claim stream now leans the way the experts always did.
The literature moved house
Refereed journal output has stayed in the tens of works per decade since 2001, while 2,741 indexed works landed in 2021-2026.5 against 369 in the whole prior decade. New claims now live before the referee, not after.
Every chart on this page is drawn from the register ledger and the archived census files. No estimates, no trackers. Raw counts are always the marks; the one derived series, the five-year average on the claims timeline, is drawn beside the raw bars and labelled, never instead of them. When the ledger changes the page is regenerated, and every figure keeps its receipt beside it.