| … | … | @@ -99,7 +99,10 @@ comes from the world (data / selection), never from the dice. |
| 99 | 99 | |
| 100 | 100 | ### The critical honesty |
| 101 | 101 | |
| 102 | | **No rung ever emerged from the rung below.** Each was a *separate model*, and the "ladder" was a |
| 102 | **No rung ever emerged from the rung below.** *(Qualified 2026-07-31 — see Part D: in the weak |
| 103 | form, one rung now does. `chain12.c` measures an autocatalytic set preferentially sustaining |
| 104 | self-templating species, 1.25× against a permutation null at chance, replicated over 484 networks. |
| 105 | That is enrichment, not open-ended heredity, and the rest of this paragraph stands.)* Each was a *separate model*, and the "ladder" was a |
| 103 | 106 | story told across disconnected pieces. The genuine emergences — rung 1 (RAF), rung 4 |
| 104 | 107 | (compartmentalization), `evoworld` (world-model) — are **isolated islands**. The self-building chain |
| 105 | 108 | 1→2→3→4 was **never achieved**. Rungs 2, 3, 0, and `flagella` were equations solved and narrated as |
| … | … | @@ -866,6 +869,63 @@ Honest bound on the linkage result: 2→3 is the rung where privatization was *l |
| 866 | 869 | begin with. The principle it demonstrates is real; the rung was the easy one; and it does not |
| 867 | 870 | transfer to 1→2 on its own — only Law 2 does. |
| 868 | 871 | |
| 872 | --- |
| 873 | |
| 874 | # Part D — one substrate, and the first cross-rung result (2026-07-31) |
| 875 | |
| 876 | Part C ended with rungs still living in different models. The cold read found that was structural, |
| 877 | not accidental: **every rung had a different substrate** — bit strings, count vectors, reaction sets, |
| 878 | gene-type lists, role counts — so a rung's output was never expressible as the next rung's input, and |
| 879 | the ladder could only ever be narrated across disconnected pieces. |
| 880 | |
| 881 | **The type.** A reaction fires only when a specific *set* of species is present — reactants and a |
| 882 | catalyst. Combinatorial by construction, which is the property Part C's failures demanded: `genome.c`, |
| 883 | `replicase.c` and `replisome.c` all failed because nothing was expensive enough to be worth |
| 884 | remembering, while `linkage.c` — the one that worked — required a **combination**. The recursive move |
| 885 | is that **a container is itself a species**. |
| 886 | |
| 887 | | milestone | result | |
| 888 | |---|---| |
| 889 | | `substrate.c` | rungs 1 & 2 coexisting; reproduces `rung1.c`'s threshold exactly (0.3/0.25 at p=0.006); gives `polymer.c`'s bootstrap problem free (maxRAF 0 alone, 1 with a seed) | |
| 890 | | `container.c` | containers are ordinary species (products, nestable) — **but presence/absence cannot express rung 4**: RAF is monotone, and 0 of 1200 partitions held closure the mixture lacked | |
| 891 | | `counts.c` | abundances added; **rung 4 expressed**, recovering `ribozyme.c`'s ordering *and* `byproduct.c`'s isolation control (compartments without selection 0.526–0.535 vs well-mixed 0.536) | |
| 892 | | `ladder35.c` | **rungs 3 & 5**; error catastrophe crosses at μ=0.162 against Eigen's analytical ln(σ)/L = 0.173 | |
| 893 | |
| 894 | **Five independent prior results fall out of one representation** — `rung1.c`'s percolation threshold, |
| 895 | `polymer.c`'s bootstrap paradox, `ribozyme.c`'s compartment ordering, `byproduct.c`'s |
| 896 | isolation-≠-compartmentalisation, and Eigen's error threshold to within 0.011. The last is the |
| 897 | strongest evidence the type is real rather than merely general: ln(σ)/L is analytic, and a substrate |
| 898 | built for autocatalytic sets lands on it without being told. |
| 899 | |
| 900 | ## `chain12.c` — does rung 1 produce rung 2? |
| 901 | |
| 902 | Askable, for the first time, of one system. A species is **self-templating** when a reaction produces |
| 903 | it and it catalyses that same reaction. Are self-templaters preferentially *sustained* by a RAF? |
| 904 | |
| 905 | The control is a **permutation test**: both arms share the seed, so network and dynamics are |
| 906 | byte-identical — only which species are *labelled* self-templating differs, with the same number of |
| 907 | self-catalysing reactions and therefore the same firing advantage. |
| 908 | |
| 909 | | PCAT | networks w/ self-templater | REAL | SCRAMBLED | difference | |
| 910 | |---|---|---|---|---| |
| 911 | | 0.002 | 51 / 200 | 1.270× | 1.033× | +0.237 | |
| 912 | | 0.004 | 92 / 200 | 1.280× | 1.002× | +0.278 | |
| 913 | | 0.008 | 155 / 200 | 1.235× | 1.026× | +0.209 | |
| 914 | | 0.016 | 186 / 200 | 1.231× | 1.020× | +0.211 | |
| 915 | |
| 916 | **Held 4 of 4, 484 networks**, ~1.25× across an 8× density range, control pinned at chance. The first |
| 917 | run gave 1.433× at n=25; replication shrank it to ~1.25× — what an overestimated small-n result does, |
| 918 | and it stayed real. Conditioning is visible rather than hidden: networks with no self-templater are |
| 919 | excluded from *both* arms, since the control uses the same networks. |
| 920 | |
| 921 | **What it is not.** The weak form. Enrichment is not open-ended heredity; `template.c`/`templceil.c` |
| 922 | showed that needs faithful modular copying and `polymer.c` showed it cannot self-start at |
| 923 | un-catalysed fidelity — none of it touched. A bias in what is sustained, not a bootstrap. Rung 1→2 is |
| 924 | **not crossed**. |
| 925 | |
| 926 | **What it is.** The first replicated cross-rung result in the arc, and it survived the same |
| 927 | matched-condition check that retracted three other claims the same day. |
| 928 | |
| 869 | 929 | ## What is proven vs open |
| 870 | 930 | |
| 871 | 931 | **Established here:** |
| … | … | @@ -919,6 +979,12 @@ transfer to 1→2 on its own — only Law 2 does. |
| 919 | 979 | which is an argument for *why* the genotype/phenotype split was necessary rather than an |
| 920 | 980 | observation that it happened. ✓ |
| 921 | 981 | |
| 982 | - **One substrate expresses all five rungs** (Part D) — and independently recovers five prior |
| 983 | results, including Eigen's ln(σ)/L to within 0.011. ✓ |
| 984 | - **Rung 1 biases the system toward rung 2** (`chain12.c`) — 1.25× enrichment of self-templating |
| 985 | species against a permutation null at chance, replicated over 484 networks and an 8× density |
| 986 | sweep. The **weak form**: enrichment, not heredity; a bias, not a bootstrap. ✓ |
| 987 | |
| 922 | 988 | **Open (field-wide, not just here):** |
| 923 | 989 | - **rung 1→2 (metabolism → heredity)** — genuinely unsolved; our null echoes it. `template.c` isolates |
| 924 | 990 | the missing ingredient (a faithful modular copy/template) but does *not* show it arising unaided. |
| … | … | @@ -984,6 +1050,9 @@ Part B: `lawspace.c` `rung1.c` `rung2.c` `rung3.c` `evoworld.c` `rung34.c` `rung |
| 984 | 1050 | Part C (boundary kind, 2026-07-31): `coupling.c` `partition.c` `signal.c` `linkage.c` |
| 985 | 1051 | `byproduct.c` `chemoton.c` `rafcomp.c`; |
| 986 | 1052 | Φ_r probes `phir.h` (shared estimator) `ribophi.c` `rung34phi.c` `gardphi.c`. |
| 1053 | Part D (substrate, 2026-07-31): `substrate.c` `container.c` `counts.c` `ladder35.c` `chain12.c`; |
| 1054 | rung 4->5 under Law 2: `rung45irr.c`; genome thread: `genome.c` `replicase.c` `replisome.c` |
| 1055 | `fidfloor.c` `byproduct.c` `chemoton.c` `rafcomp.c`. |
| 987 | 1056 | Agent line: `belief_field_agent.c` `agent_learn.c`. |
| 988 | 1057 | `src/`: original Phase-1 filesystem active-perception agent (active > random search). |
| 989 | 1058 | |