The project is not a collection of experiments. It is a chain. Each link carries a specific claim, a specific measurement, and a specific way it can be wrong. The chain begins in the simplest possible world — a grid of self-referential units with no observer — and ends at the largest possible scale — the structural collapse of the post-war international order. Bacteria. RNA. Bach. UN. Four substrates. One mechanism. Two operations. Merge what is similar. Prune what is not reinforced. The claim at every link is the same: the instrument reads structure without being told what structure is. The falsification at every link is specific: the structure it reads does not survive the control. The chain is not a proof. It is a sequence of tests, each one capable of breaking it. So far, none have.
| Link | Substrate | What the instrument reads | Measurement | Falsification |
|---|---|---|---|---|
| Bacteria | 64-unit spatial grid | Self-organized boundary emergence without observer | L4/L6 frame density; spatial gradient correlation | Shuffle neighbor connections → boundary differentiation vanishes |
| RNA | Biological sequence (A/C/G/U) | Translation grammar — codon-level operational constraint | Shuffle hierarchy; CDS-UTR cross-harm delta | Dinucleotide-preserving shuffle → cross-harm signal unchanged |
| WTC | Musical sequence (MIDI) | Tonal convergence through cross-Self harm and Codex inheritance | Codex dimension distribution across pieces; cross-Self harm divergence | Random-piece Codex → convergence pattern identical to real |
| UN | Diplomatic decisions (ideal points) | Great-power alignment fracture; structural collapse precursor | Annual displacement ranking; P5 ablation; forward prediction | 2028 disp returns below 0.20 → prediction failed |
Bacteria. Self-referential structure emerges without a designer. The architecture does not need to be told where the boundary is. It finds the boundary by hitting it. This closes the GEME-BGM arc: the static prism breathed, and the breath organized itself.
RNA. The same mechanism reads operational structure in a substrate that was never produced by a mind. The genetic code was not designed. It was compressed by four billion years of replication, mutation, and selection. The instrument reads its grammar — codon organization, translation constraint — without knowing the code exists. This extends the arc from cognitive to physical: externalization is not something minds invented. It is something the universe does.
WTC. Externalized Codex changes what the next generation can see. Generation 1 processes the clean stream and precipitates centroids. Generation 2 inherits the Codex and, when the stream is perturbed by noise, queries the inherited memory to stabilize its processing. The known is buffered. Attention is released. The third generation sees what neither of the first two could see. This is the main EE proof: Codex operation — the active loop where inherited memory shapes ongoing processing.
UN. The instrument issues a falsifiable forward prediction at civilizational scale. The 2025 great-power alignment displacement is a seventy-nine-year record. The precursor pattern matches the Soviet collapse of 1991 and the Ukraine war of 2022. The prediction is time-bounded (2026-2028), operationally defined (three verification criteria, any two confirm), and specific about what failure means (the first time in 79 years that extreme displacement did not precede structural collapse). This is not a proof. It is a test. The answer arrives within three years.
The chain is not four independent experiments that converge on the same conclusion. It is a single argument unfolded across four substrates.
Bacteria proves the architecture self-organizes. RNA proves it reads structure in a pre-cognitive substrate. WTC proves externalized structure changes cognition. UN proves the reading scales to the largest human systems.
Each link depends on the previous one. If the bacteria grid had not produced spontaneous L4 emergence, the architecture would have failed at the first test — it would not breathe. If the RNA shuffle hierarchy had not separated CDS order from dinucleotide bias, the architecture would have failed at the second test — it would not read non-cognitive structure. If the WTC Codex had not converged to different tonal centers for different pieces, the architecture would have failed at the third test — externalization would not change what the next generation sees. If the UN prediction fails — if 2028 disp returns below 0.20 without a structural collapse — the architecture fails at the fourth test: it does not scale.
Four links. Four tests. Each one capable of breaking the chain. So far, none have. The next one — UN 2026-2028 — is the first that the future will judge.
Every mechanism in this chain was first tested as a scaffold — a temporary injection that does manually what the architecture should eventually do natively. The bacteria experiment had no scaffolds. It was the only link that ran natively from the start — because the mechanism was the simplest: merge, prune, breathe.
The RNA experiment has no scaffolds. The dual-encoding cross-harm is native — two Geruons, shared BiasField, per-window centroid tracking. The frame assay is native. The shuffle hierarchy is native.
The WTC Codex experiment uses current scaffolds. Codex lookup and generational consolidation are manually triggered. The scaffolds proved the mechanism produces meaningful signal (tonal convergence across pieces). Their removal — making Codex query and consolidation endogenous to the Self layer — is the next milestone.
The UN prediction has no scaffolds. The annual ideal-point pipeline is native — encode, process_vec, read displacement. The P5 ablation, the FRED negative control, the quarterly bootstrap, the forward prediction — all run on the same Geruon pipeline with frozen parameters.
The chain is cleanest at the links that require the least injection. Bacteria was native. RNA is native. UN is native. WTC is scaffolded. The integrity of the chain depends on removing the WTC scaffolds before claiming the chain is closed.
The chain is not equally strong at every link. Bacteria is proven and closed. RNA has strong shuffle-hierarchy and CDS-UTR evidence, but TE correlation is weak. WTC has directionally correct Codex convergence but is scaffolded and currently N=2. UN has a clean signal, P5 ablation, and a forward prediction — but the prediction has not yet been tested.
The honest posture: the chain holds from bacteria through RNA to UN. WTC is the link under construction. When the WTC scaffolds are removed and the 48-piece full run replicates the tonal convergence, the chain is closed. When 2028 arrives, the UN link is either confirmed or falsified. Either outcome is knowledge.