Timeline
Every dated event in the program, newest first. The right-hand column of each row names the document, commit, or receipt that proves the date, because a chronology assembled from memory is a story and one assembled from records is evidence.
The program began July 3, 2026.
2026
- milestone
Two new public code-and-data repositories published, each backing one of two new Zenodo records
Tyorden/plesiohedron-census (created 2026-09-04) and Tyorden/phylogenetic-generators-level6 (created 2026-09-07); both cited by DOI on the evidence ledger.
evidence: github.com/Tyorden/plesiohedron-census; github.com/Tyorden/phylogenetic-generators-level6; PROGRAM_LEDGER.md, entries dated 2026-09-04 23:06 PDT and 2026-09-07 00:10 PDT
- submitted
Three workshop papers submitted to double-blind venues
Reviews are due from the author 2026-09-20; notifications are expected 2026-09-29. Per this site’s publishing policy, a double-blind paper carries no title, venue, or PDF anywhere on this site until notification.
evidence: PROGRAM_LEDGER.md, entry dated 2026-09-06 01:22 PDT (three OpenReview submission confirmations)
- correction
Retracted a claim that a plesiohedron catalog was the only machine-readable list of its kind
The claim was made without a completed external search. A 2026 Wolfram Cloud database attributed to Ed Pegg, independently assembling what appears to be the same 2016 printed survey’s representative points, is cited as related work instead; a direct fetch of the notebook returned a service outage and reading it remains pending. Wording everywhere now states the catalog as not matched against the records checked as of a given date, never as the only list or as new to science.
evidence: paper_prep/MINT_plesiohedron/references/pegg_2026_wolfram_fetch_2026-09-05.md; PROGRAM_LEDGER.md, entries dated 2026-09-05 00:29 PDT and 00:31 PDT (own-goal 57)
- submitted
Opened a second pull request cross-referencing seven more OEIS sequences into a public Erdős-problems database
Seven one-line cross-references, for Erdős problems 308, 634, 939, 942, 943, 1140, and 1161. Open and under review as of 2026-09-11; not merged.
evidence: github.com/teorth/erdosproblems pull 399; PROGRAM_LEDGER.md, entry dated 2026-09-05 00:12 PDT
- killed
Both pre-registered record attempts ended under their own kill rules; no record claimed, no record found
T2, cyclic four-colour Ramsey R(3,4,4,4): search ceiling n = 276 against a live bound of 381, witness re-verified; T1, [99,20]_5 code: ceiling 49 equals the live bound, 50 refuted exactly. No record claimed; no record found.
evidence: paper_prep/OPEN_TARGETS_codetables_ramsey/RESULTS_2026-09-11.md (both attempts closed 2026-09-05 under their pre-registered kill rules; booked to this record 2026-09-11, own-goal 63)
- result
Phase 2 of the plesiohedron sweep certified 14 tetragonal and 151 hexagonal-family space-filling cells, none named, none matched against the records checked as of 2026-09-04
Both batches reproduced every printed row of the 2016 survey for their crystal systems first: 1,476 tetragonal rows, and 958 trigonal and hexagonal rows that a second reader then re-keyed blind with zero discrepancies. Certified means the exact tiling, symmetry and counting certificates pass; it does not mean new.
evidence: PROGRAM_LEDGER.md, entries dated 2026-09-04 (tetragonal G4; hexagonal-family G4; blind re-key accepted); MINT_plesiohedron/harness/G4_PHASE2_RESULTS.md and G4_PHASE2_HEX_RESULTS.md
- result
All 165 certified phase-2 cells were classified by exact perturbation under a pre-registered scheme: 115 open, 41 wall, 9 one-sided
The heuristic labels used at triage disagreed with the computed verdict on 35 of 149 cells, in both directions, and were retired. Open means type-stable along each tested direction, not stable on a neighbourhood.
evidence: PROGRAM_LEDGER.md, entry dated 2026-09-04 (open/wall classification accepted); MINT_plesiohedron/harness/phase2/WALL_OPEN_PHASE2.md
- milestone
A machine-readable catalog of 1,583 exact-confirmed plesiohedron types, 177 of them certified, was reconciled against all 3,315 printed rows of the 2016 survey
The three digitized table sets sum to the survey’s own stated totals, an end-to-end check on the transcription, and an independent recount sharing no code with the builder reports zero failures. To the program’s knowledge the first such catalog; no DOI yet, and the venue is undecided.
evidence: PROGRAM_LEDGER.md, entry dated 2026-09-04 (catalog v2 accepted); MINT_plesiohedron/catalog/STATUS.md and RECONCILIATION.md
- result
Two polyhex sequences stalled since July at n = 19 were computed to n = 32 by four independent determinations; staged for the OEIS, not yet proposed
A258205(32) = 808,490 and A258204(32) = 750,157,868,462. The way in was to count only the mirror-symmetric shapes and recover the rest by subtraction from a published column; for a third of the new indices an identity fixes the value from published data alone.
evidence: PROGRAM_LEDGER.md, entry dated 2026-09-04 (S2 accepted); S2_polyhex_symmetry/results/SYMMETRIC_RESULTS.md
- correction
The August explanation of the July polyhex failure was found wrong: the true cause had been diagnosed in the project’s own notes on 2026-07-10 and not read
Two programs had agreed on wrong values because they shared one canonicalization step, which is what the submitted methods proposal says; the later "pinched boundaries" explanation was the error, recorded as the program’s own (own-goal 50).
evidence: PROGRAM_LEDGER.md, entry dated 2026-09-04 (own-goal 50); S2_polyhex_symmetry/notes/VERIFICATION_FABLE.md (2026-07-10)
- started
Two record attempts launched under a pre-registered harness, one on a GF(5) code and one on a four-colour cyclic Ramsey bound; no record claimed, decision date 2026-09-10
The harness first showed the listed code target could not be verified on this machine in under about five CPU-years and re-aimed at a verifiable neighbour; three exact checkers agreed on 208 published codes before launch, and the controls caught three defects in the search code first.
evidence: PROGRAM_LEDGER.md, entries dated 2026-09-04 (harness built, gated, launched; accepted); OPEN_TARGETS_codetables_ramsey/RUN_PLAN.md
- external
Flammenkamp updated his octal tables and has pushed every one of the twelve games extended in July further; the July credit line stands
His corrected-mask rerun of one game landed in the same update. Every octal record on this site is therefore dated to July 2026 and none is stated in the present tense.
evidence: track_c_octal/FLAMMENKAMP_WATCH_CLOSE_2026-09-04.md (live fetch 2026-09-04; page footer dated 2026-09-03)
- correction
The program corrected its running OEIS tally from 34 to 33 after a live recount found one entry counted twice
A202182 had been counted both as an extension and as a correction. This site’s registry had derived 33 from its itemized splits since 2026-08-29; a submitted proposal still says thirty-four and is recorded in the program’s error ledger for correction at the first editorial touchpoint.
evidence: PROGRAM_LEDGER.md, entry dated 2026-09-03 (OEIS count correction); this registry’s recounts of 2026-08-29 and 2026-09-04
- result
A307185 gained a computed level-6 term, 5,539,121, from two independently written canonical forms agreeing byte for byte; not yet proposed to the OEIS
Both programs reproduced every published smaller level down to the individual generator sets before the new level was trusted. The block on this computation had been recorded as a missing compiler; the toolchain was on the machine all along.
evidence: PROGRAM_LEDGER.md, entry dated 2026-09-03 (FS01 a(6)); FS01_level6_generators/FS01_RUN_2026-09-03.md
- milestone
A workshop paper cleared its pre-submission battery: byte-identical to the frozen draft, live database diff clean, anonymized artifact built
Pending submission as of 2026-09-04, deadline 2026-09-07 early morning Pacific. The venue reviews double-blind, so the paper is not named here until its timeline allows; pending submission is not submitted.
evidence: PROGRAM_LEDGER.md, entries dated 2026-09-03 (battery) and 2026-09-04 (final readiness audit)
- decision
The census-extension paper was rejected after review at its second journal, and the reframe for resubmission began the same day
Read as a venue-fit rejection. A methods finding raised the same day, a fallback code path that never fired and was mislabelled, was verified against the code two days later, and the paper was reframed around how the indices behave as size grows.
evidence: PROGRAM_LEDGER.md, entries dated 2026-09-01 (FS04 rejected) and 2026-09-03 (draft v4 accepted)
- milestone
The cubic plesiohedron sweep closed at seven certified finalists, and two were named the Satchelhedron and the Ordenhedron, not matched against the records checked as of 2026-09-01
An exact-arithmetic sweep of all 36 cubic space groups: 1,597 orbits, 95 types unmatched against the catalog snapshot, 11 fully certified, 7 surviving every printed representative of the 2016 reference survey. Survival of the printed record is evidence of absence from the records checked, not proof of novelty; the deposit and the paper are pending, and later review found the first of the two sits on a wall of its symmetry line that the survey’s grid provably never sampled.
evidence: PROGRAM_LEDGER.md, entries dated 2026-09-01 (G5 complete; named and packaged) and 2026-09-03 (shapes paper rounds 1 and 2); MINT_plesiohedron/NAMING_DECISION_BRIEF_2026-09-01.md
- submitted
Submitted a verification-methods proposal built from the program’s own audit archive
A proposal, not a full article, to a data-science journal; the response window runs about four to six weeks and a rejection there is final. Submitted is not accepted.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-29 (manuscript number assigned)
- accepted
A398957 published, completing the Josehedron polyform family in the OEIS
The editor round asked for the process narration to come out and the verified enumerator to go up as a file, and both happened before publication.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-27
- milestone
evidence: revision log of content/results/josehedron-polyforms.mdx
- external
The census authors asked to receive the manuscript, and it went to them with the errata section flagged for their comments
evidence: PROGRAM_LEDGER.md, entries dated 2026-08-26 and 2026-08-27
- submitted
Returned the revised polyform manuscript to the Journal of Integer Sequences
The revision carries finite tiling certificates and an independent fresh reproduction; it now awaits the referee.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-25
- submitted
Submitted the metrics paper after a full author rewrite of the draft
The rewrite was prose-only under the standing numbers-verbatim rule, and it resolved the generative-AI declaration honestly.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-23
- decision
The workshop rejected the agent-workflow paper
One weak accept, one expert reject on venue fit. The reviews were saved verbatim and the material folds into a broader verification-methods piece instead of a standalone resubmission.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-22
- submitted
The census-extension paper, desk-rejected the day before, was resubmitted to a second chemistry journal and passed its technical check
evidence: PROGRAM_LEDGER.md, entries dated 2026-08-22
- milestone
Five artifact repositories archived on Zenodo with DOIs in one batch
Code, data and certificates for five projects, each with its own DOI; the deposits are the public record behind the papers.
evidence: PROGRAM_LEDGER.md, entries dated 2026-08-22; zenodo.org records API
- external
The polyform paper came back with an invited major revision, the referee having compiled the code and reproduced the counts
The referee reproduced the counts through size 8; the report’s objections mapped almost one to one onto work already staged.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-21
- submitted
Submitted the census-extension paper and sent the original authors a courtesy note the same day
evidence: PROGRAM_LEDGER.md, entries dated 2026-08-17
- correction
A sixth error documented in the census paper being extended, this one found by reading rather than computing
Five errata had been established by exact recomputation of the published tables; the sixth is a contradiction between the paper’s prose and its own table.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-16
- accepted
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-15 (afternoon)
- result
The benzenoid census extension completed at both new sizes, each merged total matching the published sequence exactly
The run gate required the merged totals to equal the published A018190 values before any table shipped, and both did.
evidence: PROGRAM_LEDGER.md, entries dated 2026-08-15 (anchor 2, census completion, evening)
- result
evidence: frontmatter of content/results/josehedron-polyforms.mdx
- milestone
Written the day A398958 and A398959 published; A398957 (fixed class) still in review.
evidence: revision log of content/results/josehedron-polyforms.mdx
- milestone
A field survey produced fifteen prepared targets, and every agent found errors in it
Nothing was executed: documentation, verified citations, correctness anchors and pre-registered kill criteria only. Three fabricated numeric tables were caught by pulling the raw sources.
evidence: paper_prep/FS00_INDEX.md
- accepted
Terence Tao merged the Erdős-problem contribution into his crowdsourced database
PR #375 to the erdosproblems repository: the OEIS cross-reference for problem 993, citing the exhaustive tree verification to n = 32. A reviewed database contribution, merged by the repository owner; not an endorsement of a proof.
evidence: github.com/teorth/erdosproblems PR #375; CAREER_MATERIALS_MASTER_2026-07-17.md, addendum dated 2026-08-07
- result
evidence: frontmatter of content/pages/about.mdx
- result
evidence: frontmatter of content/results/example-entry.mdx
- milestone
Example entry added when the site was built.
evidence: revision log of content/results/example-entry.mdx
- accepted
The eleven-entry octal campaign closed: the last three entries published in one day
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-05
- decision
Parked a well-known open problem after finding the field already crowded
Prior work had already sieved far past any naive approach, one claimed proof had been publicly refuted, and four people were listed as actively working on it. No compute was spent.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-05
- result
The tree enumeration completed at the next size and matched the published sequence exactly
Zero violations across sixteen independent chunks.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-04
- submitted
Opened a pull request linking a problem in a public problem collection to its sequence
Counts triple-checked against the published b-file, with an AI-assistance disclosure per that repository’s policy.
evidence: PROGRAM_LEDGER.md, entry dated 2026-08-04
- result
Extended a Sidon-set sequence and verified it against every published term
Zero mismatches against the published terms, plus an independent re-verification of the underlying values.
evidence: paper_prep/OPEN_TARGETS_runs/sidon_a143824/
- decision
Started an append-only ledger so nothing in the program can be quietly lost
Status changes get a new line. Dead items stay listed and marked dead.
evidence: PROGRAM_LEDGER.md:1-9
- result
Finished an Erdős-problem tree enumeration and posted it publicly
Earlier work by another contributor reached a smaller bound in April 2026 and is credited alongside it.
evidence: OPEN_TARGETS_runs/RESULTS.md; repo Tyorden/erdos-993-trees-n31
- external
The extended octal data is confirmed live on Flammenkamp’s canonical reference page, credited by name
The previous record holder had already reviewed the solver and reproduced the flagship computation; this is the public listing the announcement was waiting on, screenshot-confirmed in the July audit. His page lists eleven games; the fleet work covered twelve, and public statements about the page use his count.
evidence: Tyler_Orden_Master_Content_Bank_v16.md §0 (v16 audit, screenshot-confirmed) and §VI-4
- submitted
Submitted the pricing paper to a journal with no author-side charge
evidence: PROGRAM_LEDGER.md:14
- correction
An audit of a two-and-a-half-day window found a missed discovery in already-shipped output
One new rare value in a game whose run had been recorded as finding none.
evidence: AuditArchive/MODEL_AUDIT_2026-07-16.md:24-25
- correction
The same audit found an overnight database check had been offset-blind
All ten games mapped to existing entries, so the work was extensions rather than new sequences. Corrected before anything was submitted.
evidence: AuditArchive/MODEL_AUDIT_2026-07-16.md:72-87
- decision
A journal desk-rejected the metrics paper
evidence: track_g_metrics/BUILD_LOG_WSJF.md:12-14
- killed
An adversarial check confirmed the rejected paper had no novel theorem to save
Its central result was a 1956 theorem. The finding was that pivoting to another venue would have been dishonest rather than unlucky.
evidence: track_g_metrics/NOVELTY_CHECK_2026-07-15.md:10, :17-25
- external
The previous record holder checked the octal solver himself and confirmed the results
evidence: AuditArchive/MODEL_AUDIT_2026-07-16.md:29
- correction
A dropped digit that had sat in a published sequence for fourteen years was corrected
Established by four independent methods before submission. The original author replied: "great, thanks!"
evidence: MASTER_SUBMISSION_GUIDE.md:18-19
- external
The OEIS board updated its AI policy to be tools-neutral
evidence: MASTER_SUBMISSION_GUIDE.md:46-56
- correction
Caught a request-only note in staged submission material before it went live
Never reached the database. Found by reading the staged text rather than trusting it.
evidence: MASTER_SUBMISSION_GUIDE.md:34-45
- milestone
Swept every subagent audit report from eight projects into one archive
The audit record now survives sessions across the whole program rather than living in one repository.
evidence: AuditArchive/MANIFEST.md:3
- external
An OEIS editor rejected three b-file edits on a formatting rule, and was right
B-files must not mirror the DATA field. The uploads were withdrawn and redone.
evidence: MASTER_SUBMISSION_GUIDE.md:117-119, :128-131
- submitted
Submitted the agent-workflow paper to its workshop, three days before the deadline
evidence: OPERATION_LOG.md:256-257
- accepted
Three new sequences published, with their b-files uploaded the same day
evidence: OPERATION_LOG.md:260-262
- submitted
Submitted the polyform paper to the Journal of Integer Sequences
evidence: MASTER_SUBMISSION_GUIDE.md:134
- external
Reported the octal record to the researcher who had held it
evidence: MASTER_SUBMISSION_GUIDE.md:136
- accepted
The polyform extension was published in the OEIS about an hour after it was proposed
evidence: OPERATION_LOG.md:98
- result
Computed the octal game 0.007 to a bound four times the one standing since 2021
Seven hours on a single desktop core. Not the current frontier: see the entry for what has happened since.
evidence: track_c_octal/RECORD_RESULTS.md:3, :9-12
- result
Proved a hendecahedral honeycomb unique, then counted its polyforms for the first time
Nobody had computed a single value, because nobody had written down how the cells fit together.
evidence: track_e_followup/sphenoid_results.md:3
- correction
A six-agent audit found the source data corrupt, not the analysis
Nine episodes in the television corpus carried the wrong transcript, and a third of one series was truncated at a spreadsheet cell limit. Only a comparison against the source could have caught it.
evidence: SVU/docs/PAPER_BUILD_LOG.md:16-21
- correction
Re-ran the whole television analysis on the corrected corpus. No finding flipped
evidence: SVU/docs/svu_paper_stats_v2.md:3; commits 4afebd1, d596bc9
- started
Started looking for a publishable result that fits in under an hour of compute
Three tracks launched in parallel: continued exponentials, polyform enumeration, and octal games.
evidence: OPERATION_LOG.md:3-18
- killed
Killed the continued-exponential track at the novelty gate, the same day it started
The convergence-domain result was already stated in the abstract of a 1996 paper. The gate runs before the compute for exactly this reason.
evidence: OPERATION_LOG.md:28-32; track_a_continued_exp/novelty_gate_report.md
- result
Found four new terms for a polyform sequence nobody had extended since its creation
evidence: OPERATION_LOG.md:49-58
- result
Re-derived one of those terms with a second, independent algorithm before submitting
A different implementation in a different language, agreeing exactly.
evidence: OPERATION_LOG.md:66-70