| … | … | @@ -42,6 +42,14 @@ a boundary out of concatenation: linked genes *cannot* replicate at different ra |
| 42 | 42 | not on the menu, and it emerges un-forced from independent replicators. Law 2 explains the whole of |
| 43 | 43 | Part B in one line: supplied compartments worked because the boundary was **free**. |
| 44 | 44 | |
| 45 | **Refined by Part C's last two experiments.** Law 2 is about the *boundary*, and `byproduct.c` shows |
| 46 | that is only half of what a compartment is: total isolation, supplied free, achieves nothing (5.65) |
| 47 | until compartments also **divide and are selected** (10.78). And `chemoton.c` shows that second half |
| 48 | is *derivable* — couple growth to the internal chemistry's own output and compartment-level |
| 49 | differential reproduction follows with no decree. So the operative principle is broader than Law 2: |
| 50 | **a boundary does work when it is a consequence of the chemistry rather than an investment in it** — |
| 51 | whether by concatenation (2→3), precipitation, or growth-coupled division. |
| 52 | |
| 45 | 53 | --- |
| 46 | 54 | |
| 47 | 55 | ## Part A — the world model (learning) |
| … | … | @@ -481,7 +489,7 @@ boundaries were **spatial**. That exhausts the options in chemistry; it does not |
| 481 | 489 | digital substrate, where individuality is routinely established by neither containment nor |
| 482 | 490 | connectivity — a writer set, a keypair, a signature. So Part C sweeps the *kind* of boundary. |
| 483 | 491 | |
| 484 | | All four use chemistry identical to `ribozyme.c` (φ = hairpin fold, parasites 3× faster, F0=0.85, |
| 492 | All six use chemistry identical to `ribozyme.c` (φ = hairpin fold, parasites 3× faster, F0=0.85, |
| 485 | 493 | F1=0.999, 4800 units, 200 gens, start φ≈5.99 which is also the neutral expectation of 6.0), with |
| 486 | 494 | **no compartments, no space, no group selection, global Wright-Fisher** — so they are directly |
| 487 | 495 | comparable to a known failure: ribozyme well-mixed goes 5.99 → 5.50. |
| … | … | @@ -493,6 +501,8 @@ comparable to a known failure: ribozyme well-mixed goes 5.99 → 5.50. |
| 493 | 501 | | **topology** (breakable graph) | `partition.c` | partitioning and gating **anti-correlated** | |
| 494 | 502 | | **verification** (costly signal) | `signal.c` | **first rise above neutral** — but invadable | |
| 495 | 503 | | **concatenation** (linkage) | `linkage.c` | **EMERGES** — the crossing | |
| 504 | | **precipitation** (byproduct wall) | `byproduct.c` | free and non-discretionary, and **the wrong half** | |
| 505 | | **growth-coupled division** | `chemoton.c` | **compartment-level selection emerges** — no decree | |
| 496 | 506 | |
| 497 | 507 | ### `coupling.c` — the gap-junction reframe |
| 498 | 508 | |
| … | … | @@ -625,17 +635,96 @@ Separately worth recording: the paper computes Φ_r **inside a single supplied a |
| 625 | 635 | It is a detector operating *within* an individual, and says nothing about where one comes from, |
| 626 | 636 | which is precisely our open problem. |
| 627 | 637 | |
| 638 | ### `byproduct.c` — the boundary was the wrong half |
| 639 | |
| 640 | Law 2 said the way past 1→2 is a boundary **nobody can decline to build**, so: walls precipitated as |
| 641 | an unavoidable side effect of catalyzing. Walls start at **zero**, accumulate from local activity, |
| 642 | decay, and gate mixing symmetrically. Parasites are not forbidden from depositing — they have |
| 643 | nothing to deposit. The only difference from `bridges.c` arm A (spatial, failed) is the wall. |
| 644 | |
| 645 | | deposition | wall OFF | BYPRODUCT | SHUFFLED | SUPPLIED | **SUP + sel** | **BYP + sel** | wall mean | r | |
| 646 | |---|---|---|---|---|---|---|---|---| |
| 647 | | 0.00 | 5.56 | 5.56 | 5.58 | 5.65 | **10.78** | 6.00 | 0.000 | +0.00 | |
| 648 | | 0.10 | 5.56 | 5.59 | 5.60 | 5.65 | **10.78** | 6.03 | 0.467 | +0.48 | |
| 649 | | 0.20 | 5.56 | 5.63 | 5.63 | 5.65 | **10.78** | **11.34** | 0.937 | +0.57 | |
| 650 | | 0.40 | 5.56 | 5.66 | 5.66 | 5.65 | **10.78** | **11.39** | 1.000 | +0.00 | |
| 651 | |
| 652 | B5 passed (walls formed, varied, tracked activity r ≈ +0.5) and B1 passed (spatial arm falls). **But |
| 653 | the pre-registered positive control B2 failed, and its failure is the finding:** supplied walls at |
| 654 | full height — total isolation — reach only **5.65**, no better than open space. Adding between-site |
| 655 | selection takes supplied walls to **10.78** and byproduct walls to **11.39**. |
| 656 | |
| 657 | > **Isolation is not compartmentalization.** The barrier was never the active ingredient. What made |
| 658 | > compartments work is that they **divide and are selected** on catalytic content. |
| 659 | |
| 660 | Law 2 survives but addresses **the wrong half**. A compartment is isolation *plus* differential |
| 661 | reproduction; byproduct precipitation buys the first half free and non-discretionarily, exactly as |
| 662 | predicted, and leaves untouched the half that does the work. Secondary and real: *given* the second |
| 663 | level, byproduct walls **beat** permanently supplied ones (11.39 vs 10.78) — self-sealing where |
| 664 | activity is high while low-activity regions re-mix recycles failed regions better than a permanent |
| 665 | box. Byproduct walls are a good compartment; they are not a source of the selection level. |
| 666 | |
| 667 | This also reframed the standing question. It was never *how do you pay for the boundary* — boundaries |
| 668 | are cheap and available as a byproduct. It is **where does differential reproduction of compartments |
| 669 | come from, when individual-level selection favours the parasites?** |
| 670 | |
| 671 | ### `chemoton.c` — and that turns out to be derivable |
| 672 | |
| 673 | A compartment divides when it grows big enough, so differential reproduction reduces to differential |
| 674 | **growth** — and if the wall material comes from the chemistry inside, growth is automatically a |
| 675 | function of catalytic content: |
| 676 | |
| 677 | > catalysis → material uptake → growth → division |
| 678 | |
| 679 | Plus the ingredient `byproduct.c` lacked: **material is conserved** (pool + stores + units×cost = |
| 680 | 6000, verified exactly in every cell) and rationed when demand exceeds supply, so compartments |
| 681 | genuinely compete. Decay without competition is not selection. **No group-selection step exists in |
| 682 | the code** — nothing samples parents by fitness. The within-compartment tragedy is untouched: |
| 683 | parasites still replicate 3× faster and still take the growth slots. |
| 684 | |
| 685 | | division size | **COUPLED** | CONTROL (uptake blind to φ) | comps | mean size | div/gen | `ribozyme.c` (decreed) | |
| 686 | |---|---|---|---|---|---|---| |
| 687 | | 4 | **11.68** | 6.36 | 2317 | 1.87 | 100.7 | 11.64 | |
| 688 | | 8 | **11.45** | 6.00 | 1397 | 3.41 | 23.4 | 11.34 | |
| 689 | | 16 | 7.65 | 5.76 | 798 | 6.34 | 5.7 | 5.90 | |
| 690 | | 32 | 5.86 | 5.62 | 426 | 12.25 | 1.2 | 5.73 | |
| 691 | |
| 692 | **C5 passed** — real turnover, so compartment-level reproduction existed to select on. **C4 passed** — |
| 693 | the control, with uptake constant per unit and blind to φ, stays at or below neutral in *every* row: |
| 694 | division-plus-assortment alone does nothing, and the growth→catalysis coupling does the work. |
| 695 | **C2 passed** — φ 5.99 → 11.68. |
| 696 | |
| 697 | **C3, the key test, is PARTIAL by its own pre-registered bar.** Three of four points land on the |
| 698 | decreed version almost exactly (11.68/11.64, 11.45/11.34, 5.86/5.73), but division size 16 gives |
| 699 | **7.65** where the decree gave 5.90 — a soft shoulder rather than a cliff, scoring small 2/2, large |
| 700 | 1/2. Caveat offered as a caveat, not a rescue: compartments here die back continuously, so realized |
| 701 | mean size is roughly *half* the threshold (1.87 … 12.25), which puts "division size 16" at ~6-unit |
| 702 | compartments, inside ribozyme's working range. The verdict stays PARTIAL. |
| 703 | |
| 704 | **This is the first time in the arc that compartment-level differential reproduction arose instead |
| 705 | of being supplied**, and it reproduces the decree's *signature* — crossing at small division sizes, |
| 706 | failing at large — rather than merely producing some rise. It is Gánti's chemoton coupling, the |
| 707 | field's standing answer; the contribution is testing it against our own measured claim that isolation |
| 708 | is insufficient, and finding that it supplies exactly the half isolation could not. |
| 709 | |
| 628 | 710 | ### Where Part C leaves rung 1→2 |
| 629 | 711 | |
| 630 | | No closer, and **narrowed**. Wiring, topology and verification are now closed off as substitutes for |
| 631 | | a compartment, and Law 2 says *why*. The ladder's asymmetry has one explanation: **boundary |
| 632 | | formation emerges when, and only when, the boundary costs nothing to defend** — 3→4 had free |
| 633 | | between-compartment selection, 2→3 has free concatenation, 1→2 has neither. |
| 712 | **Narrowed to one thing.** The circle that stood since the 14-experiment arc — privatization needs |
| 713 | group selection, group selection needs a compartment, a compartment is what we were trying to derive |
| 714 | — is broken at a specific joint: `chemoton.c` derives compartment-level selection from growth |
| 715 | coupling, with no decree. What remains is **not** a boundary problem at all. |
| 716 | |
| 717 | What still does not exist is a **template**. `chemoton.c` is the fidelity/replicase half of |
| 718 | `ribozyme.c`: φ is a *supplied* fold→catalysis map, and what rises is catalytic quality, not |
| 719 | open-ended heredity. The compartment now builds and reproduces itself; the genetic template still |
| 720 | does not. Per `template.c` and `templceil.c`, unlimited heredity needs templated copying, and per |
| 721 | `polymer.c` that cannot self-start at un-catalyzed fidelity — Eigen's bootstrap paradox, untouched by |
| 722 | anything in Part C. |
| 634 | 723 | |
| 635 | | The next candidate follows from Law 2 rather than from another guess: a **byproduct compartment**, |
| 636 | | where chemistry builds its container as an unavoidable side effect of catalyzing (mineral |
| 637 | | precipitation, self-templating pores). Wall-building is then **non-discretionary** — you build iff |
| 638 | | you catalyze — so it cannot be defected on the way σ was. |
| 724 | So the ladder's asymmetry has a cleaner explanation than the one Part C opened with: **boundary |
| 725 | formation emerges wherever the boundary is a consequence of the chemistry rather than an investment |
| 726 | in it** — 3→4 free between-compartment selection, 2→3 free concatenation, and now 1+4 via growth |
| 727 | coupling. What does not follow from any of it is a faithful modular copy. |
| 639 | 728 | |
| 640 | 729 | Honest bound on the linkage result: 2→3 is the rung where privatization was *least* contested to |
| 641 | 730 | begin with. The principle it demonstrates is real; the rung was the easy one; and it does not |
| … | … | @@ -661,6 +750,11 @@ transfer to 1→2 on its own — only Law 2 does. |
| 661 | 750 | - **Rung 2→3 (independent replicators → linkage) emerges.** ✓ — the easy rung, but genuinely tested. |
| 662 | 751 | - **A boundary made of edges does not substitute for one made of walls** (`partition.c`) — |
| 663 | 752 | partitioning and gating are anti-correlated, so exclusion and mutual access cannot be had at once. ✓ |
| 753 | - **Isolation is not compartmentalization** (`byproduct.c`) — supplied total isolation achieves |
| 754 | nothing (5.65) without differential reproduction of the compartments (10.78). ✓ |
| 755 | - **Compartment-level selection need not be supplied** (`chemoton.c`) — coupling growth to the |
| 756 | chemistry's own output generates it with no fitness-proportional sampling anywhere, reproducing the |
| 757 | decreed version's size-threshold signature (PARTIAL: small 2/2, large 1/2). ✓ |
| 664 | 758 | |
| 665 | 759 | **Open (field-wide, not just here):** |
| 666 | 760 | - **rung 1→2 (metabolism → heredity)** — genuinely unsolved; our null echoes it. `template.c` isolates |
| … | … | @@ -670,8 +764,10 @@ transfer to 1→2 on its own — only Law 2 does. |
| 670 | 764 | not produced *by* rung 2.) |
| 671 | 765 | - **spontaneous origin of a templated/modular substrate** — we *supplied* modularity and the copy |
| 672 | 766 | primitive; whether either self-organizes from random chemistry is the deep open edge. |
| 673 | | - **a non-discretionary boundary for 1→2** — the byproduct compartment, where the wall is an |
| 674 | | unavoidable side effect of catalysis. Follows from Law 2; **unbuilt**. |
| 767 | - **the template** — the one thing left of the original wall. `chemoton.c` gives compartments that |
| 768 | build and reproduce themselves, but φ is still a *supplied* fold→catalysis map and what rises is |
| 769 | catalytic quality, not open-ended heredity. Per `polymer.c`, templated copying cannot self-start at |
| 770 | un-catalyzed fidelity (Eigen's bootstrap paradox) — untouched by anything in Part C. |
| 675 | 771 | |
| 676 | 772 | --- |
| 677 | 773 | |
| … | … | @@ -689,12 +785,13 @@ Part A: `stream.c` `sns.c` `mess3.c` `learn.c` `critical.c` `competitive.c` `da- |
| 689 | 785 | `wakesleep.c` `wakesleep2.c` `mom.c`. |
| 690 | 786 | Part B: `lawspace.c` `rung1.c` `rung2.c` `rung3.c` `evoworld.c` `rung34.c` `rung45.c` `bridge.c` |
| 691 | 787 | `rung45dup.c` `flagella.c` `gard.c` `gardenv.c` `digigard.c` `template.c` `templceil.c` `spatial.c` `window.c` `polymer.c` `ribozyme.c` `bridges.c` `coemerge.c` `integrated.c`. |
| 692 | | Part C (boundary kind, 2026-07-31): `coupling.c` `partition.c` `signal.c` `linkage.c`; |
| 788 | Part C (boundary kind, 2026-07-31): `coupling.c` `partition.c` `signal.c` `linkage.c` |
| 789 | `byproduct.c` `chemoton.c`; |
| 693 | 790 | Φ_r probes `phir.h` (shared estimator) `ribophi.c` `rung34phi.c` `gardphi.c`. |
| 694 | 791 | Agent line: `belief_field_agent.c` `agent_learn.c`. |
| 695 | 792 | `src/`: original Phase-1 filesystem active-perception agent (active > random search). |
| 696 | 793 | |
| 697 | | Per-experiment write-ups: `coupling.md` `partition.md` `signal.md` `ribophi.md` `rung34phi.md` |
| 698 | | `gardphi.md`. |
| 794 | Per-experiment write-ups: `coupling.md` `partition.md` `signal.md` `byproduct.md` `chemoton.md` |
| 795 | `ribophi.md` `rung34phi.md` `gardphi.md`. |
| 699 | 796 | |
| 700 | 797 | Each `*.c` is standalone: `cc -O2 -std=c11 -o X X.c -lm && ./X`. |