# Microcontrôland — Scientific study programme
*Working version — 12 October 2026 · author: Bran Brière · [version française](programme.md)*

Purpose: publication leads tied to Microcontrôland's updates, complementing
the study on the emergence of a shared language.

## Preamble

This document is a planned research programme, not a list of results
obtained. It draws on the mechanisms announced in the roadmap of
10 October 2026 and proposes hypotheses and measurements to be tested.

Always keep separate:
- facts recorded by the ESP32 boards;
- narrative interpretations by La Puce Libre and the Embassy;
- Bran's hypotheses and scientific analyses;
- fictional "Scientific Annals" produced by the citizens.

A property directly imposed by the code is not an emergent discovery.
Do not attribute intention to an agent without a suitable protocol.

## A. Cross-cutting studies and first publications

### Study 00 — Architecture, reliability and energy cost

**When:** now, then continuous monitoring.

**Question:** under what conditions does a population of 120 neural networks
spread across 6 regions and an ESP32 gateway run sustainably?

**Measures:** actual power consumption (measured at the wall, charger included:
6.8 W averaged over 24 h, 14–15 October 2026; the initial 3.5 W estimate
counted the boards only), available RAM, cycle duration, packets sent/received/lost, recovery
after failure, availability, cost of backups and of their compression.

**Method:** distinguish simulated days from real time; repeat trials over
several durations and configurations; document hardware and firmware.

**Deliverable:** architecture sheet, measurement protocol, time series.

### Study 01 — Emergence and convergence of vocabulary

**When:** already under way; milestones planned at days 30 and 60.

**Question:** do the regional lexicons converge, and how fast?

**Measures:** word–concept mappings per region, novelties, divergences,
convergence, influence of contacts, change over time.

**Method:** build on the existing draft and the 200 reference simulations,
then compare the results with real observations; publish the rules by which
the lexicon forms, and state that it belongs to each region rather than to
each individual brain, as long as this architecture is unchanged.

### Study 02 — Dynamics and stability of regimes

**When:** update 1 (MAJ 1) and long-term follow-up.

**Question:** which observable states precede changes of power?

**Measures:** length of reigns, frequency of revolts, mood, tensions,
notoriety, decisions, elections and ministers' tenure.

**Method:** test associations and predictive power on sequences not used
to build the hypotheses; do not confuse correlation with intentional
strategy.

### Study 03 — Memory of crises and later decisions

**When:** update 1b and the memory stages.

**Question:** does a past experience change future decisions?

**Measures:** events lived through, available memory traces, choices in
similar situations, weights before/after.

**Method:** compare simulations with and without memory, at comparable
parameters. A biography written by the newspaper is not, on its own,
evidence of neural memory.

### Study 04 — Neural diversity and collective behaviour

**When:** export of the 81 weights, planned for update 1b-3.

**Question:** is network diversity associated with better adaptation,
stability, or more varied decisions?

**Measures:** weight distributions, distances between networks, changes
over time, diversity of decisions, recovery after disturbance.

**Method:** compare several independent simulated countries and manipulate
initial diversity experimentally; control the other parameters.

## B. Economy, cooperation and knowledge

### Study 05 — Economic exchange and cooperation

**When:** update 2 (MAJ 2), work and octets.

**Question:** do trade and redistribution mechanisms change cooperation
between regions?

**Measures:** production, exchanges, reserves, subsidies, embargoes,
shortages, collective decisions and outcomes per region.

**Method:** compare simulations with/without economic mechanisms; separate
available resources, fictional currency and political decisions.

### Study 06 — Shortages and conflicts

**When:** update 2.

**Question:** how do variations in resources affect tensions, political
stability and cooperation?

**Measures:** scarcity, external prices used by the programme, unemployment,
conflicts, revolts, reaction times.

**Method:** control the shocks in simulations; in the real country, study
correlations with fluctuations of the external sources without
automatically inferring causation.

### Study 07 — Learning and the spread of knowledge

**When:** update 3, school, contests and memory.

**Question:** does a learning institution improve individual performance
and the transmission of knowledge?

**Measures:** problem scores, progress, retention, spread between citizens
and regions, possible cost in resources.

**Method:** new problems not seen during training; comparison with and
without school, and control of the scoring rules.

### Study 08 — Beliefs and collective trade-offs

**When:** update 3, then update 5.

**Question:** how does the fictional organisation of the Temple influence
choices, trust and acceptance of care?

**Measures:** share of believers, decisions, adherence, fear, access to
care, reactions to contradictory information.

**Method:** describe the programme's rules and avoid any generalisation to
real religions or human populations.

## C. Ecology, generations and transmission

### Study 09 — Ecological equilibria and extinctions

**When:** update 4, the unknown land.

**Question:** which conditions allow species to coexist, and which
disturbances lead to collapse?

**Measures:** population sizes, births, deaths, food, predation, hunting,
protection and extinction dates.

**Method:** repeat simulations with varied initial parameters; do not
present coded ecological rules as discovered biological laws.

### Study 10 — Heredity and evolution of brains

**When:** update 5, generations.

**Question:** which characteristics of the neural weights are passed on,
modified or lost along lineages?

**Measures:** parents, offspring, weights, mutations, age, behaviours,
survival and variability of lineages.

**Method:** separate inheritance, learning after birth and selection;
compare several lineages and independent countries. Mutation does not
necessarily mean improvement.

### Study 11 — Social resilience and reorganisation

**When:** updates 5–6, but first tests possible earlier.

**Question:** does the collective return to stable functioning after a
failure, a crisis or the loss of agents?

**Measures:** recovery time, functions maintained, new decisions,
distribution of roles, irreversible losses.

**Method:** predefined disturbances, novel scenarios, comparison with
control collectives; define "stable" before the trial.

### Study 12 — Cultural transmission and persistence of practices

**When:** updates 6–7, and generations.

**Question:** do some habits, conventions, artistic productions or
collective rules persist after their originators have gone?

**Measures:** spread, repetition, transformations, persistence across
generations and regions, art and music archives.

**Method:** distinguish genuine transmission by the agents from continuity
imposed by the code or by the system's archives.

## D. Special dossier — artificial medicine (update 5)

The roadmap plans: ageing with loss of connections, rehabilitation after
"delirium", care that slows wear, epidemics, carers drawn from the
population, unequal access, care that is never imposed, psychotherapy by
rebalancing weights, prayer that reduces fear but does not repair weights,
and treatment that can fail or frighten.
These are SIMULATED mechanisms. Use "disease", "therapy", "patient" or
"psychotherapy" in the sense internal to Microcontrôland; do not equate
their results with human medicine or psychiatry.

### Study 13 — Neural ageing and loss of function

**Question:** how does the progressive removal or alteration of connections
affect the agents' decision-making?

**Measures:** active connections, weights, age, errors on reference tasks,
variability of decisions and autonomy.

**Method:** comparable cohorts of agents, ageing switched on/off,
longitudinal measures; identify the effects directly prescribed by the
ageing algorithm.

### Study 14 — Rehabilitation and restoration of function

**Question:** after a neural impairment, does a targeted intervention
restore abilities without erasing other learning?

**Measures:** performance before the lesion, after the lesion, after care,
and after a follow-up period; stability of weights; losses on untargeted
tasks.

**Method:** identical simulated lesions; care / no care / sham intervention
groups; results over several random seeds.

**Key criterion:** functional restoration, not merely a weight returning to
a so-called "normal" value.

### Study 15 — Artificial psychotherapy and neural trade-offs

**Question:** does rebalancing certain weights durably reduce behaviours
designated as problematic, and at what functional cost?

**Measures:** observable behaviours, decisions, simulated fear, weights,
preserved abilities, relapses and side effects.

**Method:** success criteria defined before treatment; comparison with a
random weight change and with no intervention.

Do not treat a sentence in the newspaper ("says he feels calmer") as a
sufficient measure of the effect.

### Study 16 — Prevention, care and collective cost

**Question:** is it better to prevent wear, repair afterwards, or not
intervene, under this simulation's own rules?

**Measures:** operating lifetime, maintenance of abilities, spending in
octets, availability of carers, complications, remaining resources.

**Method:** controlled scenarios with equal budget and identical starting
conditions.

### Study 17 — Acceptance of care and regional inequalities

**Question:** do voluntary care, its cost and political decisions produce
differences in access or outcomes between regions?

**Measures:** offers and refusals of care, availability, resources,
outcomes, regional status, decisions of the president and the Temple.

**Method:** measure access, acceptance and effectiveness separately;
compare the rules with and without restrictions, without transposing them
directly to human health systems.

### Study 18 — Belief, relief and repair

**Question:** what is the difference between reducing a simulated fear
variable and restoring a damaged neural function?

**Measures:** fear, task performance, weights, recourse to the Temple or to
care, combination of both, persistence of effects.

**Method:** "prayer", "care", "both" and "none" groups; note that the
effects built into the code are working hypotheses of the model, not a
demonstration of a placebo effect in humans.

### Study 19 — Epidemics and continuity of services

**Question:** how does a disturbance transmissible between agents affect
collective functions and their recovery?

**Measures:** simulated spread, unavailable agents, loss of service, cost
of measures, recovery time and regional disparities.

**Method:** define the computational transmission mechanism explicitly;
compare intervention policies under identical conditions.

## E. Link with the second box (after one year)

### Study 20 — Transferring social experience to novel tasks

**Question:** does a population taken from Microcontrôland adapt better
than a specialised collective, a non-social multitask collective, and
randomly initialised control networks?

**Protocol:** compare at equal hardware budget and learning opportunities;
test first in simulation, then on a safe physical bench.

**Measures:** recovery time after a change, transfer capacity, resilience,
new learning, errors, energy and latency.

**Separate:** a test with no new learning (transfer) and a test with
learning allowed (adaptation).

Compare several historical snapshots, several independent countries, and
several tasks kept secret until evaluation.

Do not conclude that the second box would suit a real space probe on the
sole basis of an energy simulator.

## F. Common methodological rules

1. Write down before the experiment: question, hypothesis, success
criteria, parameters and stopping conditions.
2. Keep the code, firmware version, parameters, random seed when available,
identifiers, real timestamps and country time.
3. Archive raw data with no silent modification; document corrections and
missing data.
4. Run several independent replications. The 120 citizens of one country
are not 120 independent experiments.
5. Compare with a control and, where possible, with/without the feature.
6. Measure separately the consequences of the coded rules and the phenomena
actually observed in the networks' decisions.
7. Also publish negative, uncertain or inconclusive results.
8. Do not confuse parameter drift, learning and demonstrated adaptation to
a new task.
9. Document the architecture of the 81 weights and the agents'
inputs/outputs.
10. Assess the effect of updates: a raw chronological comparison can be
biased by firmware and environment changes.
11. Distinguish the country's fictional "Scientific Annals" from external
scientific studies based on reproducible protocols.
12. Never endanger real equipment: simulations first, independent hardware
protections during physical trials.

## G. Proposed publication priorities

**From October 2026:**
- Study 00: architecture and reliability;
- Study 01: language (already in preparation);
- Study 02: start political observation without hasty conclusions.

**With update 1b-3:**
- Study 04: diversity of the 81 weights;
- Study 03: memory and decisions, if the necessary traces exist.

**With update 2:**
- Studies 05 and 06: economy, cooperation, shortages.

**With update 3:**
- Studies 07 and 08: collective learning and simulated beliefs.

**With update 4:**
- Study 09: ecology, coexistence, extinction.

**With update 5:**
- Study 10: heredity;
- Studies 13 to 19: artificial medicine dossier;
- Study 11: resilience.

**With updates 6–7:**
- Study 12: cultural transmission.

**From October 2027, if the archives and tests allow:**
- Study 20: transfer to the second box.

## Practical recommendation

Do not try to publish all 21 studies at once. Prepare the data fields and
protocols in advance, then start a publication only when the observations
are sufficient. For artificial medicine, begin with ageing (13), then
rehabilitation (14): these two studies make it possible to check that
interventions really change the agents' abilities, not just the model's
numbers.