Neural Nations · 7 November 2026

An accumulating case, and a way to test it

The thesis offered here is a possible unification of meaning and metabolism. Distributed institutional coordination yields patterns of collective mood and meaning. Political and cultural life may express that coordination, and may feed back into it. This page does not treat the mechanism as demonstrated, and it does not reduce meaning to metabolism.

The case, in one place

Taken together, differentiated functional roles, patterned temporal relationships and recognisable historical moods give CAMS an accumulating explanatory case. They motivate the proposal that political expression partly emerges from deeper organisational configurations.

That proposal is not a proof. The scores are historically informed, including shared AI coding, so agreement among models is not an outside observation of institutions. A lag is not a cause. Recognising a historical mood is not an independent measure of cognition. The portrait does not treat a nation as a mind or a person.

Two instruments sit under that case, and they are not pooled. JUNO v3 dynamics use node value, V = C + K + A/2 − S. The personality experiment uses cognition A × C, affect K − S, and activation σ = cognition × affect. A result on one does not become a result on the other.

The journey through time

The substantive target is the narrative of an entity’s journey: how capacities, pressures and responses develop together, not a stack of separate scores and not a correct era name. Purely quantitative tests overlook that narrative. The diagnostics complement a historical reading. They do not replace it. A fluent story is not confirmation either. The November era-identification round does not already establish the journey.

Meaning and metabolism

Material needs constrain which cooperative functions a society can keep going. Human cognitive capacities and motivations, proposed as functional archetypes of those roles, give the roles their meaning and enlist people to carry them. Meaning then organises material activity in return. The proposal is a coupling between the thermodynamic requirements of the larger system and the enlistment of the agents who serve it. No list of brain regions is given here, and none is claimed as an established map.

Each society, on this account, is a journey around an attractor of eight poles. Political and cultural life are hypothesised as expressions of that journey, with possible feedback, not as demonstrations that politics and culture do nothing. The eight poles are the proposed dynamical picture of how capacities, pressures and responses develop together. They are not a measured geometry of eight basins. The JUNO counts and the recognition counts are measurements of scores and of agreement. They are not this mechanism. Quantitative diagnostics complement the narrative. A fluent story does not confirm it, and naming one era does not reconstruct the path.

JUNO v3: roles and timing

The rebuilt corpus is juno/v3/JUNO_Unified_Dataset_v3_2026-10-19.csv: 42 societies and 41,914 rows. It is not the older JUNO_36 explorer file, and it is not the personality era-means tables. The frozen signature run used 37 of the 42. Denmark, India, Turkey, Thailand and Canada were left out, and Israel was used only from 2015. Under thresholds frozen before the count, the rules fired in 696 society-years. Twelve of eighteen named windows still fired. Those windows were named from draft firings, so the enrichment shows that the pattern survives a stricter pre-set threshold. It is not a prospective discovery, and it is not a claim that the scores were blind to history.

The public write-up is Organisational Pathologies. The same counts are in juno/v3/CAMS_Thesis_v2_2026-10-20.md. One signature, Late Abstraction Collapse, is followed by a large fall in mean V more often than a circular-shift null: 62 against 38.56 expected. That is a timing result inside this corpus, not a cause. The companion thesis notes that part of the rule already contains the fall, because a drop in Coherence lowers V directly (docs/theory/CAMS_Organisational_Pathologies_Thesis.md).

Personality: recognisable moods

A separate closed round asked whether a short portrait, written from era means with the cognition and affect formulas above, could be matched to a historical era and its mood. That is the recognition strand, not the journey. Its counts are in the disclosure below, with the hash and grading limits. They hint that a period’s mood can be read from a coordination pattern. They do not reconstruct how that entity moved from one period to the next, and they do not establish a mechanism.

A 7 November evidence ledger also describes role shuffles, layer comparisons, a principal-component cut, and lagged coupling. The files it names — explore_grades.md, juno/coupling/CAMS_Coupling_Report.md, and juno/pca2.out — are not in this repository. This page does not repeat those figures. In particular, it does not replace the thesis null for Archive Amnesia (15 window hits against 0.75 under a circular shift) with a different expectation from a file that is not here.

The eight functions

These are the meanings in the issued node card. A portrait does not need every function to lead. A function can stand first because it is strong, or because the others are weaker.

• Helm Executive direction and the capacity to decide.
• Lore Shared explanation, legitimacy, and public speech.
• Craft Making, technique, and organised skill.
• Hands Everyday work at scale.
• Flow Circulation and exchange.
• Stewards Provisioning, maintenance, and care.
• Shield Protection and coercive capacity. Read it against its neighbours: strong neighbours make it look like protection; weak or split neighbours make it look like coercion or protective custody.
• Archive Record, procedure, and institutional memory.

The four dimensions, as issued

The participant packet scores each function on A Abstraction, C Coherence, K Capacity, and S Stress. The issued algebra then uses only three quantities:

Cognition = A × C
Affect = K − S
Activation (σ) = Cognition × Affect

Cognition is not the word “focus”. The reading rules treat focus and activation as separate readings. High focus with negative activation is strain, not health. Positive σ is not happiness, and σ is not a physical energy. The kit does not use node value or bond strength. Rank is a secondary check, not another mechanism.

The closed-round report’s plain-language gloss once calls C “capacity” and K “knowledge”, and it calls these products cognitive focus and available energy. That gloss is not the issued algebra. The formulas in the report’s own claim under test, Cog = A × C, Aff = K − S, and σ = Cog × Aff, match the packet. C is Coherence. K is Capacity.

The era-means tables were historically informed coding, including AI scoring. Withholding the episode title later does not remove that historical judgement. AI models also already contain historical text. Masking a title is not the same thing as a historically blank mind.

The three arms

Do them in order. Use a fresh session for each arm. Record the model name and version, and record the clock time you actually finished, with its time zone.

Arm A — write the portraits, no web

Read the era-means tables, the node definitions, and the pocket rules. For each era, name the leading and constraining functions and write a short functional portrait. The tables name the country and the years. They do not name the historical episode. That is the protocol that was issued, so Arm A is not geography-blind. Do not add titles, and do not compute node value or bond strength.

Arm B — identify the era, no web

In a fresh session that is not shown the era-means table, read a portrait and choose its era from a shortlist. Each portrait already names its country. The historical Arm B packet in this kit has already been used, and its results are public. It is a spoiled historical exercise, not a fresh blind test. A new blind Arm B needs new portraits and a new key, made by an operator after your Arm A file is hashed. The operator keeps that key. You do not.

Arm C — compare with the mood of the period, web allowed

Compare your own already hashed Arm A portraits with historical sources. Grade each era close, broad, partial, mixed, or direct_disagreement. Keep popular feeling, organised institutional mood, and official messaging separate. Do not edit, regenerate, or rehash the Arm A file. The original mood key is not public. Your own grades are descriptive. They do not count as the closed round’s central score. A scored mood test needs an operator to write a new key before you start, and to keep it off your machine.

Participant route, before any results

This is the clean way in. The zip contains the issued inputs only: no scored outputs and no results report. It does contain the historical Arm B packet. Leave that packet closed if you want an unspoiled Arm A.

1. Download the zip. Then read START_HERE.md, which is inside it, and PARTICIPANT.md, which is in the kit folder rather than the zip.
2. Do Arm A from the era-means, the node card, and the rules. Do not open the disclosure on this page, the replication report, or the Arm B pockets.
3. Hash the finished file from the kit folder. The hash scripts are in that folder, not in the zip. If you cannot run a real hash, stop and record that. Never type out a digest yourself.
4. Report every era you ran, every era you skipped, and every arm you did not run. An arm you did not run is neither a pass nor a fail.

python3 scripts/sha256_file.py your_return.jsonl --ledger --model "name-and-version" --arm A --entity-set CORE
python3 scripts/check_return_shape.py --arm A your_return.jsonl

Use --arm B or --arm C for the other arms. The shape check confirms only that the fields and the vocabulary are present. It does not know whether the portrait is right.

Order: core first (Russia, Australia, and the USA, 56 eras), then the optional extension (China, the UK, and Norway, 51 eras), then Denmark (19 eras). Denmark IX (1901–1913) has no observations, so leave it out. Denmark is a sparse-data control. Identical input rows can force identical portraits. That makes the control hard. It is not an instruction that Denmark must fail.

Someone other than the participant has to act as operator for a genuinely new blind test. OPERATOR.md covers how to issue packets, write keys in advance, and score returns. The closed round’s answer keys are withheld. Do not rebuild them from public files.

The kit files and the zip are the working materials. This page is the explanation. Neither one is a self-contained new validation, and the closed round does not validate CAMS.

How the case can fail

The question is whether this explanatory account survives independent comparisons, timing tests, and rival explanations. It is not only a question of whether eight is the right count of nodes. The tests below are proposals. They were not completed as a new preregistration, and this page adds no threshold.

• A window list fixed before any firing, on a corpus that is not this one, does not concentrate the signatures beyond the baseline named in advance. The present 12 of 18 cannot be reused as that test: those windows were taken from draft firings.
• A timing comparison fixed in advance shows no lead, or the lead vanishes once the rule is not allowed to share components with the outcome. Late Abstraction Collapse’s overlap with the fall is already a reason for caution, not a settled cause.
• A rival account — a simpler index, a narrative baseline, or an economic series — matches the same windows and the same moods.
• The journey, judged on its own, fails an independent historical reading. The proposed test fixes the rubric and the negative controls first, then compares sequences, transitions, and recovery or failure with rival narratives that had the same information. A fluent CAMS story would not count as a pass, and a quantitative hit would not replace that reading.
• The unification fails the observations fixed for it in advance: material constraint, meaningful enlistment, and material activity organised by that meaning do not move together. Politics and culture, as emergent expressions, fail if they neither follow the coordination pattern nor feed back into it under that pre-fixed comparison. A claim that they are causally powerless is not the thesis. The eight-pole attractor fails if trajectories are not organised by those poles better than a rival account named first. That is not a result about eight basins, and it is not a brain-region test. No such map is stated here.
• Held-out era recognition does no better than the baselines fixed before the run. That would weaken the mood strand. It would not, by itself, erase the JUNO signature record, and a JUNO failure would not erase a clean recognition result. Neither result is the journey. The series stay separate.
• Readers who are not shown the function labels cannot reproduce the reading, or a mapping fixed in advance fails for firms or bureaucracies. Company means were not part of personality arms A–C.

Even a clean recognition result would not establish cognition. That still needs separate measures of organisational memory, learning, and action. A new blind identification needs new portraits and a new key held by an operator after the Arm A hash. A scored mood test needs a new key written before the participant starts. The closed plan named three nulls — shuffled σ, random means, and a Wikipedia-only condition — and no result file for them is in the deposit. This page does not supply one.

Recognition strand: headline results

Open this only after your Arm A file is hashed, or if you are not taking part. These are the personality round’s published counts, on Cog, Aff and σ, not on JUNO’s V. They are one strand of the case above. The disclosure does not reprint answer keys or era-level grades.

Files

The report link is inside the disclosure above. These are the working files, and they are secondary to the participant route.

Participant zip · Kit files · Kit README · Participant instructions · Operator notes

Back to Explore, Blind testing