The register

Every marked claim, including the ones taken back.

Thirty-one live claims and three formal withdrawals, pulled from the five published papers and the unpublished drafts underneath them. A withdrawn row is not a deletion — it is kept on the record, with the reason and, where it exists, what replaced it. That is deliberate: a programme that quietly fixes its mistakes is not more credible than one that publishes them.

checked against the original source from memory, not yet checked estimate novel — mine, and untested someone else's, cited

Reference hygiene

Twenty-four sources were verified at their origin rather than trusted from a citation. That pass found four wrong first authors — Tjhung→Kenchel, Powner→Singh, Aida→Ando, Bzdok→Stanley — plus a citation error on Pizzarello C-2/C-3, several framing errors, and over forty findings that changed the drafts before publication. The error count is not embarrassing; a verification pass that found nothing would be evidence that no verification pass happened. Full account: /method/.

Every row is sourced to a chain and a paper. JS adds the filter chips and column sort above; the full table is here regardless.
Mark Paper Basis
Entropy production rate is the housekeeping cost of a nonequilibrium steady state. Chain 1 21991606 Oono & Paniconi 1998; Hatano & Sasa 2001.
The Lyapunov exponent measures whether nearby trajectories converge or diverge; its relation to entropy production is open. Chain 1 21991606 Both quantities defined and explicitly distinguished, not conflated.
Four fields — neuroscience, machine learning, physiology, statistical physics — independently converge on the same frozen / critical / chaotic structure. Chain 1 21991606 Beggs & Plenz 2003; Poole et al. 2016; Cohen et al. 2021; Barato & Seifert 2015.
Tension buys fidelity, at a rate you can write down. Chain 1 21991606 Thermodynamic uncertainty relation (Barato & Seifert 2015); kinetic proofreading (Hopfield 1974; Ninio 1975).
χ is not a temperature — the escape law descends from weak-noise large-deviation asymptotics, and no thermal bath is required anywhere. Chain 1 21991606 Freidlin–Wentzell theory; Santolin, Freitas, Esposito & Falasco, Phys. Rev. E 2025.
Three regimes exist where the temperature substitution genuinely breaks: non-unique effective temperature, the drive's correlation time, non-Gaussian jump forcing. Chain 1 21991606 Stated as the published note's own validity conditions.
Annealing and canalisation are indistinguishable from inside the system — same escape law, one moves ε and one moves ΔΦ. Chain 1 broad paper, unwritten Gradient descent = escape law at T=0; canalisation raises ΔΦ instead of raising T.
The inversion method is only valid under slow recovery — quench fast and it returns a confident wrong answer. Chain 1 broad paper, unwritten Falls out of the ε/ΔΦ separation as a validity condition, not an assumption.
Any account of the genetic code's origin must not select for translation before translation exists. Chain 2 22008798 The field's own stated gap: Crick 1968; Nissen et al. 2000; Wolf & Koonin 2007.
The code is the frozen interface contract of a merger between two independently selected RNA systems. Chain 2 22008798 The peptidyl-transferase centre as generator, charged acceptor stems as carrier.
Hartman & Smith 2019 established operational-code priority, the thioester world, CCA, and code-at-a-coupling first. Chain 2 22008798 Prior-art collision recorded and credited in §1, not absorbed or footnoted.
Hartman & Smith's coevolutionary mechanism selects components toward a coupling before it exists — the same teleological trap this account was built to avoid. Chain 2 22008798 Named as the specific mechanistic difference, not just a citation.
Chirality is transduced, not determined — local chemical bias does not fix a global sign; a freezing event does. Chain 2 22012402 Glycine is achiral, so the lock cannot predate amino acid number two — the sharpest falsifiable corner.
The Catma is the freezing event that collapsed code, amino-acid handedness and backbone handedness together. Chain 2 22019657 Distinguished from the central dogma, which is the rule once frozen — the Catma is the freezing.
The central dogma does not describe the historical order RNA→protein→DNA — it prohibits transfer out of protein and explicitly permits RNA→DNA. Chain 2 22019657 Crick's own 1970 paper exists to correct exactly this misreading.
No break in life's continuity shows — a second origin would have had to independently arrive at the same code, handedness, residue set and CCA terminus. Chain 3 bioblob draft, unpublished Chains 2.1–2.4 explain why that contract would not be arrived at twice.
Exactly one property has been selected in every lineage simultaneously: being on the other side of a discontinuity. Chain 3 bioblob draft, unpublished Not fitness, complexity or fecundity — those are selected within regimes and deleted between them.
A poly-extremophile is adapted to the transition between extremes, not to an extreme itself. Chain 3 bioblob draft, unpublished The Hadean read as a succession of lethal environments, not one difficult one.
The units-of-selection objection is a slice error — a phylogenetic tree is the internal structure of one object, drawn as though from outside it. Chain 3 bioblob draft, unpublished Selection in a whole is ordinary biology: Physarum's tube network, somatic evolution, ramets within a genet.
The Gaia comparison sharpens rather than dissolves the argument: Gaia needs selection on the whole; this needs selection in the whole. Chain 3 bioblob draft, unpublished Every tree and every body already has selection in the whole; Gaia cannot.
The totality of life is a single connected region through spacetime, because cytoplasm has only ever been produced by partitioning existing cytoplasm. Chain 3 bioblob draft, unpublished Nearest counterexample, SpudCell (Adamala lab, 2026), used externally supplied ribosomes — translation borrowed, not re-ignited.
A spore or tun is carrier-only: record held, generator and filter both stopped, and unambiguously not dead despite zero dissipation. Chain 3 bioblob draft, unpublished Conceded deliberately rather than discovered under review.
Life uses roughly 0.2% of the chemistry reachable from six simple substrates. Chain 4 Wołos et al. 2020, Science 82 biotic molecules and 36,603 abiotic ones reached by generation 7 of 614 transforms.
Biotic molecules are roughly four times more robust to reaction-rule deletion than abiotic ones on the same network. Chain 4 Wołos et al. 2020, Science 34 of 63 reaction classes removable for biotic synthesisability, vs. 8 of 63 for abiotic.
Scale-free network architecture cannot explain the robustness asymmetry, because biotic and abiotic molecules share the same graph, hubs and topology. Chain 4 framework, unpublished Wołos et al. report the asymmetry and attribute it to architecture without checking this.
The same operation that destroys information about the past is what preserves the readable part of it — destruction is selective, biased toward frozen optima. Chain 5 method register Deeply canalised things survive precisely because the lip is high.
An extinction destroys information about what was and creates information about where the minimum must be — a many-to-one map, not a loss. Chain 5 method register The inversion the whole method turns on.
Landauer's bound is true, astronomically loose, and does no work here; phase-space contraction is the usable version. Chain 5 method register Evans & Searles 2002; Gallavotti–Cohen fluctuation theorems.
Verifying 24 sources at their origin produced four wrong first authors and over forty findings that changed the drafts. Chain 5 method register Tjhung→Kenchel, Powner→Singh, Aida→Ando, Bzdok→Stanley — see the note above.
A set-valued pairwise adjacency descriptor cannot represent oligomeric closure, because it discards multiplicity and locality. Chain 5 22073289 1CR0 is annotated "hexameric" and is a 6₁ screw-filament excerpt — annotation is not ground truth.
The replacement descriptor made a blind prediction that held: 1E0J's closure, reproduced edge-for-edge with zero spurious edges. Chain 5 22073289 Established by all-atom contact before the fragment pipeline saw it.
The single-commit property of a pre-registration document makes contemporaneity machine-verifiable via git log. Chain 5 method register A small, novel observation about method itself, not about the framework's content.
Withdrawn
Compositional assemblies (composomes) as an early heritable carrier.
Chain 2 22019657 Withdrawn — see below

Withdrawn — published in The Central Catma's claims table, Aug 2026.

Reason. Vasas et al.: imposing a fitness gain on a target compositional assembly yields a relative selective advantage of 1.00715, because off-diagonal fitness-matrix elements exceed the diagonal ones — an assembly is fed more by its neighbours than by its own offspring, so it cannot hold identity long enough for selection to grip.

What's kept. Vasas et al. explicitly grant the generator half of the claim ("accumulating prebiotic material and increasing environmental patchiness"), and that half is retained. The merger (Chain 2) then isn't swapping one carrier for another — it's supplying the first carrier to a generator that never had one.

Withdrawn
The merger obliged an archive — that translation created a functional need for a DNA archive.
Chain 2 22019657 Withdrawn — see below

Withdrawn — published in The Central Catma's claims table, Aug 2026.

Reason. Retracted as teleological: the claim argued from what the merger needed rather than from what was selected for. This correction came from Meow-Ludo, not from Claude.

Replaced with. Environmental conditions, not internal needs, select for DNA.

Withdrawn
The bioblob predicts a performance-breadth trade-off in continuously-selected lineages.
Chain 3 bioblob draft, unpublished Withdrawn — see below

Withdrawn — bioblob draft v2, Sept 2026. Not yet in any public claims table before this page.

Reason. A phylogenetically informed comparison of 28 teleost species found peak aerobic capacity unlinked to thermal performance breadth, and a review of niche-breadth evolution concludes performance-breadth trade-offs may not be common determinants of niche breadth at all.

Replaced with. Preferential gene loss as the predicted signature instead. See Chain 3.

Found something wrong that isn't already listed here? Tell me where this is wrong.