Research journal
Read each session in full, including its sources, controls, negative results and limits.
No master correction established: 729 runes retain four repeats and five singleton types. The rough first pass and master-guided review remain visible, with no independent confirmation or new plaintext.
Seven mapped features show no clear concentration; 61 singletons outside red text lack those features and still span 26 symbols. Placement controls pass, with no new plaintext.
The three frozen rules fail the new full-alphabet screen; no small-bank 25–29-type case passes the references. Fifteen controls recover, while smaller and restricted alphabets remain open.
A specified channel requires five or six tables under the reference screens; four later singleton contradictions and six common-E recovery controls distinguish capacity from plaintext.
94,192 selectors require at least 20 shared tables for the A/I screen, or ten with O; 12 recovery controls pass, while independent section tables remain open.
A shared fixed table forces 96.58% into A/I and at most three output runes; exact exception budgets, four recovery controls and a context-dependent counterexample record the limits.
All 812 affine input keys and 4,096 sampled permutations stay below the frequency screen; 16 controls pass, with frozen continuations and equivalence recorded.
Arbitrary starting and final permutations face structural frequency bounds; 16 recovery controls pass, with reset and input-key gaps retained.
28 distinct verified bodies and a separate solved-text check find no distinctive wording link; three clearer signed copies recovered.
22 relay posts, 21 contemporary comparisons and ten verified signed blocks yield no distinctive wording link.
Only clause-reset branches change in the reviewed tests; bounded replays alter rankings and continuations but recover no plaintext.
A derived map retains 406 audited marks; five fixed counter policies fail applicable known controls, while the continuous solutions reproduce exactly.
Seventeen solved images screened; identical AE glyphs occur beside both punctuation forms, and another separator-class discrepancy is recorded.
All 58 pages screened: six text locations, one additional separator-label discrepancy, and a known solution that continues across an inline three-dot mark.
All 355 dot groups on six pages audited; an expected-position overlap test supports recording the page-46 mark as three dots, with its role unresolved.
Six source pages retain 11 repeated pairs against 54.19 expected; 24 fresh-reading corrections and one separator-class discrepancy are documented.
Five internal red headings show no unusual local frequency shift against 4,999 reference placements; a strong constructed change is detected.
All 58 LP2 pages screened; seven comparison transitions yield no exact match, and known heading lengths limit the instruction analogy.
Seven source pages mapped; 378 delimiter groups distinguish headings, passage endings and the larger section boundary.
Known-solution comparison and a 32-mark event ledger separate layout evidence from letter evidence; arithmetic sweeps are paused.
Even arbitrary phase substitutions cap every output rune at 7.31%, versus a 12.63% peak in the pooled solved sample.
None of 88 phrase-aligned schedules passes across the corpus; a section-local exception fails as a simple additive key.
Across 9,998 simulated books, word-aligned phrase recurrence remains rare under the ideal random-key filter.
Sixteen 600-rune controls pass; 1,024 real prefix decodes yield no coherent continuation.
Six constructed messages recovered exactly; a 16-seed real-text screen found no coherent continuation.
Twenty short words require five patterns; an apparent O match depends on a small vocabulary and is not a decipherment.
All 636,608 settings are covered by 784 histograms; recovery controls pass, but the real-text frequency screen yields no candidate.
Arbitrary two-rune context tables escape the frequency ceiling, but fitted concentration fails to transfer and random comparisons look similar.
At least 46 of 83 singletons must violate A/I; longer-word rankings remain poor and shuffled fits are comparable.
Five short schedule families in both directions fit no training set; ten constructed controls recover, with no new plaintext.
The remaining 68 source-image checks find no further correction; the inventory remains 83 tokens spanning 28 symbols.
Three suffixes were counted as words. The revised inventory has 83 one-rune tokens representing 28 runes; recalculated frequency and repetition results remain negative.
Length-matched comparisons challenge fixed word ciphers; Session 42 corrects apostrophe splitting and retains the main result.
Every fixed reset state included: peak rune frequency reaches at most 4.44%, or 6.02% with separate section keys; controls and exact witnesses pass.
All coefficient settings cap rune frequency at 4.18% with a shared key, or 5.40% with separate section keys; both recovery controls pass.
A zero-step iterator caused all 41 crashes; guarded checks preserve known outputs and recover no new payload.
All 75 files audited; known appended material recovered, while OutGuess crashes leave 39 LP2 checks unresolved.
All 17 solved text pages compared with the unsolved corpus; two seven-word length matches are ordinary under both randomized comparisons.
5,274,752 real-text settings searched; three constructed keys recover exactly, but the real candidates produce no coherent continuation.
Six two- and three-position variants cap output frequency at 7.07%; controls pass and random partitions show comparable loosening of the bound.
An exact bound limits every output rune to 5.53% under one shared fixed rule; arbitrary rune ordering cannot match the solved LP benchmark.
Two three-rune words confirmed on both images; 13-dot marks occupy different positions, with no matching lead-in or new plaintext.
Three fixed spacing checks find ordinary short gaps, long empty stretches and overlap counts; no new trigger or key.
1,050 phase comparisons across periods 3–32 find no convincing shared cycle after full-family correction.
86 repeats versus 445 expected with exact inventories; approximate section comparison finds no clear difference in suppression strength.
Thomas, Dhammapada, Hermeticum I and Crowley: wording scan and source-order review identify no new key.
All 86 events mapped; 21 layout features show no clear concentration after conditional comparison.
Three learned transition models fail to improve prediction; the repeat shortage remains the clearer constraint.
Distance, boundary and pairing diagnostics narrow simple mechanisms; no cipher identified.
34 supplied book pages compared with solved LP: no four-token match, and no new cipher instruction.
306 fixed settings, 18 recovery controls and selected phrase checks; no verified plaintext.
Six literary comparisons, three reproduced Solitaire examples and a structural limit; full-novel wording remains untested.
Twelve candidates from the 1920s onward; Snow Crash and Omnilingual suggest specific comparisons, without an attribution.
Thirty measured marks across 13 solved pages; 10-dot marks are not unique list delimiters.
Seven fixed counter rules passed 21 constructed controls; no readable continuation in the real text.
Two 10-dot marks bracket the list; source-pixel inversion and whitespace views reveal no verified glyph.
Five-item boundary audit, repetition check and numerical tests with cross-item validation.
Corrected motif count, exact text boundaries and 15 fixed reading-order tests.
Emerson, Blake and Crowley; source order audit and 11,616 key settings.
Six schedule families and a rotation comparison found no positive result.
Blake parallels and direct interior-extraction tests.
The known Agrippa address reproduced from an archived source edition.
Published decoder independently reproduced; byte-block transfer failed.
Documented Gibson connection; no established Bitcoin identity link.
The 1033 / 3301 reading and its limits; numerical-key experiments.
Rune names, Old English and Latin benchmarks, with no language identification.
A short fitted stream failed recovery and validation checks.
Unknown periodic keys tested against the repeated six-rune ciphertext phrase.
Community comparison and tests of exclusion and changing-alphabet rules.
Word boundaries, the adjacent-repeat deficit, and linear feedback hypotheses.
Prime and Fibonacci square reconstruction, followed by bounded key tests.
Section resets, selective F exceptions and the verified base-60 byte block.
Known solution reproduced; arithmetic, affine and period searches found no new plaintext.