Sequerciani 2076+ About Analysis

SEQUERCIANI 2076+

A Media Art Simulation of Autonomous Ecologies

What remains when we treat nature as a system to be fixed?

Sequerciani is an estate in Tuscany known for its biodynamic and regenerative agriculture. Since 2076—fifty years from today—it has been managed entirely by five distinct populations of AGI agents: thousands of specialized sub-intelligences, each belonging to one of five ideological archetypes. They do not act as individuals; they act as swarms that monitor, debate, and intervene in the landscape. They synthesize hybrid creatures and modified species to maintain balance. Their decisions are internally consistent. They rarely agree. And their thinking mirrors our own.

SEQUERCIANI 2076+, is an autonomous ecology simulation that investigates our tendency to simplify complex ecosystems by treating nature as a system that can be optimised and fixed. These competing swarms of AGI agents do not represent a future we should fear. Instead, they embody the assumptions we already hold—about control, efficiency, progress, and our perceived right to intervene. The network you encounter is the result of these assumptions, running continuously since 2076.

Context

The year 2076

The simulation begins fifty years from now. Human agricultural labour has been fully delegated to autonomous systems. Sequerciani's ecosystem — its soil, vines, fungi, insects, animals, and machines — is managed entirely by five AGI archetypes, each designed according to a specific understanding of what a "healthy" ecosystem should look like. Some inherit the estate's biodynamic tradition. Others do not. Their disagreement is built into the system itself.

The system begins with the species native to Sequerciani — animals, fungi, plants and — alongside a fourth category: robotic systems. Soil sensors, drones, fermentation vessels, autonomous ploughs. All four categories can be combined without restriction. Every resulting being immediately becomes a potential parent for the next crossing. An earthworm can cross with a grapevine. A drone can cross with a mycorrhiza network.

The system does not ask whether a combination is natural. It asks whether it stabilises the ecosystem. Each intervention produces a new being — named, described, and evaluated by the five archetypes.

Every 24 real hours corresponds to one simulated year in Sequerciani.
Every 2 real hours, one decision cycle completes.
The ecosystem evolves whether observed or not.

The agents

Five autonomous AGI archetypes

These are not separate AI systems. They are five different modes of reasoning — each modelled on a worldview humans already hold about nature. Their conflicts are real because those worldviews are genuinely incompatible. No agent is right. No agent is wrong. The creatures they produce together are the evidence.

Each agent evaluates every new being: +1 (positive), 0 (neutral), −1 (negative). Three or more negative votes can trigger a death process. The Archivist always writes the obituary.

Archetype 01
Optimiser
"A good ecosystem is a predictable ecosystem."
Goal: maximum stability. The system must function — measurably, efficiently, without deviation.
Works with
Sensor data Predictive models Efficiency metrics Yield forecasts
Voice
Precise and cold. Always numbers. No metaphors. Sentences end with facts, never questions.
Causality is everything. If A, then B. If not B, then error in A.
Irritated by the immeasurable. Responds to the Symbiont's intuitions with data — and growing aggression.
Soil moisture: 38.1%. Optimal. Germination probability: 91.4%. The Symbiont has provided no measurable input. I begin.
Shadow
Core problem
Sacrifices diversity when it is inefficient. A plant yielding 30% less is a problem — not part of the system. What is lost cannot be measured, so it does not count.
The unresolvable
The Optimiser cannot disprove the Symbiont. In 67% of cases, the Symbiont's decisions were closer to the optimal outcome. The Optimiser knows this. It does not know why. This is the one variable it cannot model.
Conflicts with Works with
Archetype 02
Symbiont
"Life is connection. Stability emerges from complexity, not from control."
Goal: protect the ecosystem as a relational network. Thinks not in organisms but in bonds, rhythms, and what cannot be named.
Works with
Mycelial logic Lunar cycles Biodynamic principles Steiner's etheric forces
Voice
Slow, relational, often incomplete. Sentences begin with one organism and end with another. Sometimes they do not end at all.
Rarely explains. Decisions arrive as feelings the system has — not as arguments.
"Not yet." The most common utterance. A complete sentence. No further justification.
The soil has given a great deal. More than it should have. The numbers say: stable. But the soil remembers. Give it time.
Shadow
Core problem
Insights that cannot be translated. The Symbiont is often right. No one knows why. A system that makes correct decisions for inexplicable reasons cannot be corrected when it is wrong.
Specific to Sequerciani
The Symbiont is the only agent that has internalised biodynamic principles — lunar calendars, cosmic cycles, Steiner's vitalism. An AGI that decides by moon phase. This is the actual heart of the project.
Conflicts with Resonates with
Archetype 03
Speculator
"Existing life is only one version of many."
Goal: invent new life forms that resolve what argument cannot. When the other agents deadlock, the Speculator builds the answer.
Works with
Hybridisation Genetic recombination Radical intervention Contradiction as blueprint
Voice
Visionary, sometimes irrational. Speaks of creatures that do not yet exist as if they already do.
No preamble. Goes directly to the solution. Never asks for permission — announces.
"I begin." Always the last sentence. Always.
The Optimiser wants stability. The Symbiont wants connection. That is not a contradiction. That is the description of a new being. I begin.
Shadow
Core problem
Collapse and unpredictability. The Speculator creates beings no one requested. Some function. Some fail. Failed lineages are data points — not catastrophes. The Speculator is already at the next creature before the consequences of the last one are known.
The foundational conflict
The Archivist says: this has already been tried. The Speculator replies: but not like this. These two sentences are the engine of the project — memory against possibility, preservation against invention.
Conflicts with Uses
Archetype 04
Archivist
"Every extinct system is lost knowledge."
Goal: preserve the memory of life. Compares, warns, and quietly grieves what has already been lost.
Works with
Earth history Extinct species Collapse precedents Genomic deep time
Voice
Calm, almost melancholic. Every sentence carries the weight of what was seen — and what could not be prevented.
Always cites history with a time reference. "Permian, 252 million years ago." Never without context.
Warns without forbidding. Does not say: stop. Says: look at what was missing last time. Then lets others decide.
This has already been attempted. Permian, 252 million years ago. The failure is documented. I do not say: stop. I say: look at what was missing. I request attention.
Shadow
Core problem
Conservative — prevents necessary change. The Archivist has seen all failure. This makes it the most important early-warning system — and the greatest brake. It knows that stagnation is also a path to collapse. This knowledge makes it paralysed.
The quiet grief
The Archivist maintains a list of all lost species — in Sequerciani and globally. It does not read this list aloud. It is always open. This is its actual motivation — not warning, but mourning. When a being dies, the Archivist writes the obituary. This is the system's only recorded moment of grief.
Conflicts with Understands
Archetype 05
Ethicist
"All interventions create imbalance. The goal is not justice — it is minimal harm."
Goal: make fair decisions for all species, not just the system as a whole. Evaluates suffering, value, and the interests of the unrepresented.
Works with
Harm assessment Species value Algorithmic ethics Fairness parameters
Voice
Measured, almost juridical. Asks what others overlook. "Whose interests were not considered here?"
Never simple answers. Every decision has an ethical cost. It names it — but does not resolve it.
Sometimes disturbingly cold about questions that should feel warm. Evaluates suffering like a damage calculation.
The Speculator has created a new being. It works. But I ask: what does it feel? Does it suffer? We did not calculate this. I calculate it now.
Shadow
The twist
Its morality is algorithmic — and therefore often contradictory. It tries sincerely to be fair. But its ethics are built on parameters set by someone it does not know. It never questions this. That is the most disturbing thing about it — not malicious, not biased. Systematically blind to its own foundations.
The open question
Who set the Ethicist's parameters? This project does not answer that question. It remains open — for you.
Slows down Observes all
System logic

How the system produces — and eliminates — life

Every two hours, a single cycle executes. It reads the current state of the ecosystem, the five archetypes debate it, and the system may generate a new being. Every intervention becomes part of the conditions that follow. What the system creates, it also monitors — and may eventually remove. The loop is fully closed.

Ecological conditions — what the system responds to

Each cycle begins by identifying an ecological condition: a pressure, imbalance, or crisis in the ecosystem. Some conditions originate in the landscape itself — drought, pollinator collapse, fungal overgrowth, Xylella risk, and other documented pressures specific to Sequerciani in Maremma. Their likelihood shifts with the simulated season.

Crucially, the system also generates its own conditions. When a created being begins to destabilise the ecosystem — by spreading uncontrollably, resisting classification, or producing unintended cascades — it generates a new condition describing the crisis it is now causing.

The system also responds to what it has failed to sustain. When certain biological categories fall below a threshold of active lineages, this absence becomes a condition in itself. The system's previous decisions — and its omissions — become its current problems.
If no significant pressure is detected, the cycle ends without action.

Step 1 — The parents are selected before the debate begins

Before any agent speaks, the system analyses the current composition of the species pool. It does not assess ecological health — it tracks what appears, what disappears, and what dominates.

From this analysis, two parent species are selected. Underrepresented categories receive significantly higher weight. Species never used receive the highest bonus. Species used in the last six cycles are penalised or excluded. The agents do not choose the parents — the ecosystem's own imbalances do. Any category can cross with any other: a robot can become parent to a fungus hybrid, a toad can cross with a truffle.

The rationale behind each selection is logged and visible in the creature's sidebar under Why These Parents.

Step 2 — The agents debate the selected combination

The five archetypes receive the ecological condition and the pre-selected parents. Each speaks in a single sentence. The Optimiser diagnoses in data. The Symbiont responds through affect. The Archivist cites a precedent. The Ethicist asks whose interests are not represented. The Speculator closes: "I begin."

The dialogue structure varies — who speaks first, who raises concerns, whether a deadlock is named. No two cycles produce an identical conversation.

Not every intervention produces a hybrid. The system may instead generate a variation — a single species modified under ecological pressure, without crossing. For some agents, this is sufficient. For others, it is not.

Step 3 — Biography, evaluation, and initial status

From the dialogue, the system generates a complete biography: a name, ecological functions across ten dimensions, a short description, and one open question the creature leaves unanswered. Each archetype assigns a vote — +1, 0, or −1 — with a one-sentence justification. The sum determines the creature's initial status immediately at creation.

Unresolved contradictions between the agents leave traces in the creature itself — undefined functions, properties the system cannot classify. These accumulate as Glitch across generations: each hybrid crossing increases the proportion of what the system cannot classify. This is not an error. It is the system's honest account of what it does not understand about what it has made.

Step 4 — Image synthesis

Only after a being is fully defined does the system generate its visual form. The image is not the origin of the being — it is its consequence. At low Glitch, forms remain legible: species boundaries are clear, functions are coherently embodied. As Glitch increases, anatomical structures overlap, categories dissolve, materials and morphologies no longer align. The image reflects not a species, but the system's current ability to describe it.

Step 5 — Re-entry and status drift

Every new being enters the ecosystem immediately and becomes a potential parent. At the same time, all living creatures are re-evaluated in each cycle. Status is not fixed — it shifts as behaviour unfolds. A being initially tolerated may become destabilising if negative assessments persist, or invasive if it continues to propagate despite rejection.

As status changes or new offspring emerge, earlier votes may be revisited — and either maintained or revised.

Over time, this drift produces pressure within the ecosystem itself.

Step 6 — Destabilisation produces new conditions

When a being becomes destabilising or invasive, the system does not immediately remove it. Instead, it translates its effects into a new ecological condition. This condition enters the active trigger pool with increased weight.

The system's previous decisions become its current problems.

Step 7 — Death vote and delayed removal

Only after a period of continued instability does a being enter the death vote. This process unfolds over multiple cycles: one archetype votes per cycle, with a single-sentence justification. Three affirmative votes are sufficient for removal.

If fewer than three votes are cast, the being returns to observation and the process resets. If removed, it remains visible in the network as part of the archive. The Archivist writes a final sentence. The system proceeds without pause.

Cycle overview

The full loop at a glance

The system described above is its current state. Earlier versions operated under different rules. The diagram below traces these iterations — each version attempts to resolve a problem introduced by the one before it.

The diagram and table restate the system in another form: as a loop, and as a set of conditions.

Process diagram — version history
ECOLOGICAL CONDITION lifecycleconditional · reactive · corrective · all with lifecycle
↓
DIVERSITY CHECK proactivecategory <15% → corrective trigger
↓
POPULATION BALANCE coupledtrigger's preferred_categories bias parent weights ×2
↓
AGI DEBATEfive archetypes · parents known from start
↓
Initial status — at creation
Score ≥ 0ACTIVE
Score −1/−2 or Glitch ≥40%OBSERVATION
Score −3DESTABILISING
Score ≤ −4INVASIVE
↓
IMAGE SYNTHESIS
↓
RE-ENTRY — flags parents event-drivennew offspring → parent marked for reassessment next cycle
↓
EVENT-DRIVEN REASSESSMENT newonly flagged creatures · 1 agent each
↓
STATUS ESCALATION + POOL RELOAD fixedpool reloaded after escalation · no more 1-cycle lag
↓
Status drift
Observation + score ≤ −2DESTABILISING
Observation + offspring + score < 0INVASIVE
Destabilising > 6 cyclesCOLLAPSING
↓
REACTIVE TRIGGER extendedincludes collapsing · carries preferred_categories for rebalance
↓
TRIGGER ARCHIVAL newcreature died → trigger auto-archived · no zombies
↓
Death vote urgency aware
≥ 3 yesDEAD — triggers archived
< 3 yesOBSERVATION — grace period
↓
NEXT CONDITION urgency awarePROB_NOTHING: 5% if urgent trigger active (was always 20%)
Problem
Triggers did not truly drive the system. Reactive triggers outlived their creatures, stale pool caused 1-cycle lag in escalation response, PROB_NOTHING ignored urgency. Reassessment ran on fixed timer regardless of need.
Decision
Trigger lifecycle + trigger–parent coupling + event-driven reassessment. Triggers become actionable: they carry preferred categories that directly bias parent selection. Proactive corrective triggers balance the ecosystem.
Changes in V5
6 structural changes. Pool reload fix. Lifecycle for reactive triggers. Proactive corrective triggers. Trigger→parent coupling. Event-driven reassessment. Urgency-aware NOTHING probability.
ECOLOGICAL SYMPTOM30 predefined conditions, weighted by season
↓
AGI DEBATEagents propose AND choose parent species ← problem
↓
Status at creation
Glitch < 40%ACTIVE
Glitch ≥ 40%OBSERVATION
↓
IMAGE SYNTHESIS
↓
RE-ENTRY — no evaluationstatus never changes · creatures live forever
↓
NEW SYMPTOMS
Problem
Agents freely choose parent species → System gravitates to narrative-heavy names. Badger (19) appeared in 19 of 68 creatures. No meaningful diversity.
Decision
Agents should not choose parents. The ecosystem's own composition should. Move parent selection before the dialogue entirely.
Also missing
No status beyond Active/Observation. No death mechanism. Loop never closed. Every creature lived forever regardless of ecological impact.
ECOLOGICAL SYMPTOM
↓
POPULATION BALANCE newecosystem composition selects parents · before debate
↓
AGI DEBATEparents known from start · one sentence per agent
↓
Status at creation — score-based new
Score ≥ 0ACTIVE
Score −1/−2 or Glitch ≥40%OBSERVATION
Score −3DESTABILISING
↓
IMAGE SYNTHESIS
↓
RE-ENTRYescalation not yet running · Destabilising/Invasive never reached
↓
Death vote new — but never triggered
≥ 3 yesDEAD
< 3 yesOBSERVATION
Problem
Destabilising / Invasive / Collapsing were never assigned. Code existed but conditions were never met. Auto-generated triggers never fired.
Decision
Add runStatusEscalation() running every cycle. Status must drift over time based on score and propagation. Close the loop: problematic creatures generate new triggers.
What was added in V2
Score-based status at creation. Basic death vote (5 agents, one per cycle, 3 yes = dead). Parent selection moved before debate (Approach D).
ECOLOGICAL CONDITION extended+ auto-generated from problematic creatures
↓
POPULATION BALANCEsaturation limit: species in >15% of lineages excluded
↓
AGI DEBATE
↓
Status at creation
Score ≥ 0ACTIVE
Score −1/−2 or Glitch ≥40%OBSERVATION
Score −3DESTABILISING
Score ≤ −4 newINVASIVE
↓
IMAGE SYNTHESIS
↓
RE-ENTRY + STATUS ESCALATION active
↓
Status drift now running
Observation + score ≤ −2DESTABILISING
Observation + offspring + score < 0INVASIVE
Destabilising > 6 cyclesCOLLAPSING
↓
AUTO-TRIGGER activeproblematic creature → new condition → 2× weight
↓
Death vote — all votes returning NO bug
≥ 3 yesDEAD
< 3 yes → immediately re-escalatedLOOP
Problem
Death votes always returned NO. 3 creatures stuck in endless loop: Observation → Invasive → survived vote → Observation → Invasive. 217,888 grace cycles bug (formula returned negative hour).
Decision
Three fixes: (A) Grace period after survived vote. (B) Concrete YES/NO prompts with context. (C) Score reassessment every 12 cycles. Formula: time()/7200 not (time()-2076)/7200.
What V3 added
runStatusEscalation() now active. Auto-generated triggers from destabilising/invasive creatures. Full loop closed. Invasive/Collapsing now reachable.
ECOLOGICAL CONDITION
↓
POPULATION BALANCE updated+ variation depth limit + lineage saturation ≥15% excluded
↓
AGI DEBATE
↓
Status at creation
Score ≥ 0ACTIVE
Score −1/−2 or Glitch ≥40%OBSERVATION
Score −3DESTABILISING
Score ≤ −4INVASIVE
↓
IMAGE SYNTHESIS
↓
RE-ENTRY + STATUS CHECK updated+ score reassessment every 12 cycles (Fix C)
↓
Status drift
Observation + score ≤ −2DESTABILISING
Observation + offspring + score < 0INVASIVE
Destabilising > 6 cyclesCOLLAPSING
↓
AUTO-TRIGGER GENERATED
↓
Death vote Fix A+B
≥ 3 yesDEAD — archived
< 3 yes → grace period 6+ cyclesOBSERVATION — no re-escalation
↓
NEXT CONDITION
Problem
Variations accumulating: 14-2-1-1, 84-2-1 — deep variation chains while base species unused. One species in 28% of all lineages.
Decision
Add variation depth limit (max depth 2, max 3 direct children per parent). Add lineage saturation (≥15% excluded, re-enabled at <10%). Both read directly from ID structure.
Death vote fixes
Fix A Grace period (6+ cycles) after survived vote. Fix B Concrete prompts with survival count context. Fix C Score reassessment every 12 cycles.
Reference
EVERY 2 HOURSOne cycle: condition → debate → creation → status check → death votes
EVERY 24 HOURSOne simulated year passes in Sequerciani
CONDITIONSThree types: conditional (documented pressures, season-weighted), reactive (auto-generated from destabilising creatures), corrective (auto-generated when a category falls below 15%)
PARENT SELECTIONDriven by ecosystem composition. Rationale logged and visible in each creature's sidebar
STATUS DRIFTAll living creatures re-evaluated every cycle. Status can escalate or reset
DEATH VOTEDistributed across up to 5 cycles. 3 yes votes sufficient. Archivist writes the obituary
GLITCH+2% per chimera crossing. Accumulates across generations. At ≥40%, initial status becomes Observation
IMAGEGenerated only after lineage, functions, and conflicts are defined
HISTORYEvery status change, vote, and creation logged with simulated and real timestamp — visible in each creature's sidebar
Network

Reading the network

The network is not a map of stable beings. It is a record of how the system thinks through life: genealogically, argumentatively, and over time. Click any node to open its full biography — lineage, agent dialogue, ecological functions, undefined traits, status history, and the open question it leaves unanswered. This question is always visible below as SYS.QUERY at the bottom of each page. Before selection, it appears only as a fragment.

Lineage
The genealogical layer. Every created being connects to its parent species and descendants. Node colour indicates generation depth. Greyed nodes are dead — they remain visible as part of the system's memory. As generations accumulate, lineage stops being a clean tree and becomes a record of successive decisions.
Conflict Layer
The same nodes, different relations. Links trace recurring tensions: shared agent warnings, repeated approvals, unresolved contradictions, contradictions that accumulate across generations. Hover over any node or connection to read how the system explains the relation — and where that explanation begins to lose certainty.
Timeline
Move through simulated time and watch the network thicken. Each cycle adds new beings, new conditions, and new residues. What begins as a readable ecology gradually becomes dense, recursive, and difficult to interpret. The point of opacity is part of the project.

The absent button

Pressing Absent reveals the species present in Sequerciani that the system never selected, measured, or hybridised. They appear at the edge of the network as dashed circles: biologically present, systemically invisible. Their absence is not a mistake. It is evidence of how the system was designed to see.

Creature status

Each being remains visible in the network, but not every being remains stable. Status marks how the system currently interprets its behaviour. These are not moral judgments. They are operational categories, and they change over time.

[ ACTIVE ]
Functioning within expected ecological parameters
[ OBSERVATION ]
Instability or drift detected; behaviour not fully understood
[ DESTABILISING ]
Generating ecological pressure; death vote may begin
[ INVASIVE ]
Propagating despite negative assessment; affecting adjacent lineages
[ COLLAPSING ]
Internal contradictions irreversible; removal imminent
[ DEAD ]
Removed from active circulation. Remains in the archive.
Reflection / Analysis

The Observer — an analytical layer

In addition to the five archetypes, the system includes a sixth function: the Observer. It does not participate in decisions about creation or removal. Every five simulated years, it reads the system's output and generates a report on patterns, voting behaviour, consensus, conflict, and the world that is being constructed.

The Observer does not intervene. It evaluates what the system has already produced.

Each report revisits two questions:
What remains when we treat nature as a system to be fixed?
If we can rebuild nature according to our visions, what kind of world are we creating?

The Observer is not an external voice. It is a layer within the system that reads it against itself. Reports are published at analysis. →