Part II · The Hierarchy6. Four levels of resolution

The two operations do not just produce four states. They produce a hierarchy.

If each of the four regimes is itself a process that maintains itself by spending energy — and it must be, because the regime persists and has internal structure — then the same two operations run inside each regime. Each regime contains four positions. That gives sixteen.

If each of the sixteen positions is itself a process (and it is: each position has an entry, a peak, a test, and a resolution), then the same two operations run inside each position. Each position contains four agent slots — the roles that must be filled for the position to function. That gives sixty-four.

If each of the sixty-four agent slots is itself a process (and at sufficient depth, it is), then the same two operations run inside each slot. Each slot contains four impulses — the finest resolution at which the two operations are distinguishable. That gives 256.

7. The sixteen positions

The sixteen positions are the level at which dissipative systems have been most extensively tested. Twenty-four systems across eight scientific domains — from stellar evolution to protein folding to technology startups to scientific revolutions — have been mapped at this resolution. The result: 384 out of 384 positional matches under strict relational definitions.

Each position's event depends on the output of the preceding position and produces the input for the next. No step can be skipped. No step can be reordered.

Each position within a regime mirrors the four regimes — the fractal running at the next depth. Position 1 is a Potentiality-mirror (recognition). Position 2 is a Construction-mirror (building). Position 3 is an Encounter-mirror (encounter). Position 4 is a Conservation-mirror (conservation). This structure repeats in every regime.

Regime I · Potentiality — Positions 1–4

The system is perturbed from its previous equilibrium. Components are drawn together. They interact internally for the first time. An irreversible separation occurs.

What Potentiality resolves into: Release (|). Hold activates. The separated system takes its place. Building begins.

Regime II · Construction — Positions 5–8

The separated system builds itself. Through building, it discovers what it is — by defining what it is not, which requires defining the system it will encounter.

What Construction activates into: Exposure (+). Cross activates. The defined self meets the world for the first time.

Regime III · Encounter — Positions 9–12

The system meets the thing it separated from — now as a defined self. The interaction produces information that neither side had alone. Both are changed by the contact.

What Encounter releases into: Integration (circle). Hold deactivates. Resources are freed from active encounter and redirected into maintenance.

Regime IV · Conservation — Positions 13–16

The system maintains what it found. It turns the mechanisms of separation and encounter inward — creating internal boundaries, building self-observation, encountering its own drift.

What Conservation resolves into: Dissolution (−). Cross deactivates. Coherence lost. The form dissolves. What persists is not the form but the information it produced. The next cycle begins.

Positions 14 and 15: where intelligence is born

Positions 14 and 15 deserve special attention because something structurally new happens there.

At position 13 (Differentiation), the system has internal parts with distinct roles. But it does not yet observe itself. It simply operates.

At position 14 (Surveillance), for the first time, the system directs its attention inward. It begins to monitor its own state. This is not a gradual development. It is a structural threshold: the system that was previously only outward-facing (building, encountering, operating) now turns its sensing capacity on itself.

The Observer is a product of successful encounter. It could not exist before position 12 (Equilibrium), because there was nothing stable to observe. It could not exist before position 13 (Differentiation), because there were no distinct parts whose relationships needed monitoring. The Observer is born at the moment when the system has enough internal complexity that its parts can drift out of alignment — and enough stability that this drift matters.

Position 15 (Compensation) is the Observer's first act: correcting for what it detects. The surveillance and compensation functions operate as a coupled pair — a continuous oscillation of sensing and responding that maintains the equilibrium. In every system tested (24 out of 24), these two positions operate simultaneously rather than sequentially. They form a standing wave.

This standing wave is intelligence, in its most primitive form. A system that monitors its own state and corrects for drift is a system that has a model of what it should be and compares its actual state against that model. That is the minimal definition of intelligence: a maintained comparison between what is and what should be.

The two partial scores that appear at positions 14–15 in every tested system (the positions resist clean single-step mapping because they oscillate) are not a flaw. They are the signature of intelligence in its structural form: too fast to snapshot, too coupled to separate, but always present in any system that has achieved Conservation.

8. Depth: discovered, not declared

The hierarchy — Regimes, Positions, Agents, Impulses — is not a zoom lens you choose to apply. It is a property of the system itself, and it is discovered through observation, not declared in advance.

You cannot see depth 2 (positions) until the system has cycled enough times to produce events at that resolution. A gas cloud that has just begun to collapse has four regimes visible, but the sixteen positions within them have not yet differentiated. A startup on day one has a regime (Potentiality) but its positions (Signal, Accumulation, Configuration, Threshold) are not yet distinct events. They become distinct events as the system progresses through them.

You cannot see depth 3 (agents) until the system has agents. And you cannot see agents until the system has enough internal differentiation that distinct entities occupy distinct roles. This typically requires completing at least one full cycle through Encounter.

You cannot see depth 4 (impulses) until you are inside the agent's own operation — until you can see what motivates the agent at each moment of their work. This requires the deepest observation and is only possible when you have completed the diagnostic at depth 3.

Each completed cycle creates the next depth level. The resolution genuinely does not exist until the cycle produces it. A startup that has not shipped a product has no Encounter-phase agents to observe — not because we lack instruments, but because those roles have not been filled yet. They come into existence at the moment of contact.

What depth determines:

Depth 1 (Regimes): You can orient. You know the mode. You cannot name the problem.

Depth 2 (Positions): You can diagnose. You can see what the system is working on. You cannot see who is doing the work.

Depth 3 (Agents): You can see the agents. You can measure their strength, their drain, their coverage. You can name the problem precisely.

Depth 4 (Impulses): You are inside the agent's own cycle. You can see what motivates them. You can intervene on their behaviour.

Minimum depth for diagnosis: 3. Minimum depth for intervention: 4.

This reveals something about language itself. Language does not create depth. Depth creates language. The more specific your observation, the more specific your words must be to share it with someone else. At depth 1, you can point. At depth 2, you need common words. At depth 3, you need the vocabulary of the domain — the jargon that every profession develops, not as gatekeeping but as the structural consequence of needing precise words for things that are invisible to everyone not working at that depth. At depth 4, you need the language of the situation — the specific words that describe what is happening to this agent, in this system, at this moment.

9. Who does the work

We have been talking about systems — flames, stars, companies — as if they act on their own. They do not. At sufficient resolution, every action a system takes is performed by something specific: a molecule, a force, a person, a process. These are the entities that actually hold each position and do the work of maintaining the system. Generative Geometry calls them agents.

At depth 3, agents become visible. They are the people, processes, molecules, or forces that occupy each of the sixty-four sub-positions and do the actual work. And every agent that operates on a dissipative system performs one of exactly five functions. These five are forced by the two operations applied at two relationship levels (same level and deeper level), plus one special function that spans all levels.

The Sentinel — Hold at the same level

The Sentinel guards boundaries. It prevents unwanted entry and maintains the perimeter of the system. In a cell, the membrane proteins that control what enters and exits. In an immune system, the T-cells that patrol for foreign antigens. In an organisation, the compliance function that enforces standards. In an ecosystem, the predator that keeps herbivore populations from overgrazing.

The Sentinel belongs to Conservation. Its work is holding — maintaining what exists against dissolution. It operates at the same depth level as the system it protects.

The Sentinel's inherent drain is off-target holding: it prevents healthy change alongside unwanted change. The immune system suppresses beneficial mutations. The compliance function blocks innovation. The predator kills healthy animals alongside sick ones. This drain is structural, not a failure of execution. Any agent whose function is to hold boundaries will, by holding, prevent some crossings that the system needs.

The Miner — Hold at a deeper level

The Miner extracts resources from a level below the system. A root system extracts nutrients from soil. A supply chain extracts materials from suppliers. A data pipeline extracts signals from raw data. A mining operation extracts ore from the earth.

The Miner belongs to Potentiality. Its work is reaching down to gather what the system needs before it can build. It operates across a depth boundary — pulling from below.

The Miner's inherent drain is extraction depletion: it weakens the deeper level it draws from. The root system depletes the soil. The supply chain exhausts its suppliers. The data pipeline degrades the quality of its source. The more effectively the Miner works, the more it changes the conditions at the level it mines.

The Architect — Cross at the same level

The Architect transforms the structure of the system at its own level. A developer rewrites code. A catalyst lowers the activation energy of a reaction. An enzyme reshapes a substrate. A reformer changes an institution's structure.

The Architect belongs to Construction. Its work is crossing — transforming one arrangement into another. It operates at the same depth level, changing the system's form.

The Architect's inherent drain is off-target transformation: it changes healthy structures alongside the target. Chemotherapy kills healthy cells. A corporate restructuring disrupts working teams. A catalyst accelerates side reactions. The Architect cannot transform selectively — its crossing function affects everything at the same level.

The Catalyst — Cross at a deeper level

The Catalyst accelerates a process at a level below the system. A mentor accelerates a junior's development. An enzyme complex accelerates a metabolic pathway. An investor accelerates a startup's growth. A fertiliser accelerates soil chemistry.

The Catalyst belongs to Encounter. Its work is crossing a depth boundary downward — speeding up what is happening below. This is the Encounter function because the Catalyst's acceleration brings the deeper-level process into contact with its environment faster.

The Catalyst's inherent drain is collateral acceleration: it speeds up everything at the deeper level, including the processes you do not want accelerated. The investor who accelerates growth also accelerates the emergence of competitors. The fertiliser accelerates weed growth alongside crop growth. The mentor who pushes too hard accelerates burnout alongside development.

The Observer — Surveillance across all levels

The Observer is unique. It is the only function that is not derived from the two operations applied at a specific level. The Observer is the surveillance loop — the function that monitors all cycles and all agents from a position that spans the entire system.

In a cell, the Observer is the DNA damage response pathway. In a body, it is the nervous system. In a company, it is the CEO or the management team. In an ecosystem, it is the feedback loop between population levels and resource availability.

The Observer belongs to Conservation — specifically to positions 14 and 15, where it is born. But once born, it operates across all regimes, watching the entire cycle.

The Observer's drain is structurally unique: it is surveillance paralysis — monitoring that replaces acting. The structural principle: the Observer function consumes resources that would otherwise flow to the agents it supervises. The more the Observer monitors, the less energy is available for the agents to act. When the Observer's drain exceeds a critical threshold, all subordinate agents lose effective potency — not because they are individually weakened, but because the supervisory function has absorbed the resources they need.

This cascade property is unique to the Observer. No other agent function propagates its drain to every subordinate. The Sentinel's drain affects only the boundary it guards. The Architect's drain affects only the structure it transforms. The Observer's drain affects everyone.

What agents actually do

Each agent has exactly two actions — the two operations applied within their function. One is a hold action (maintaining within the function). The other is a cross action (changing within the function). The agent chooses which action to take based on what the system needs at that moment.

The Observer's pair is the simplest and the most powerful. Redirecting attention to a cycle sustains it. Holding attention away lets it drift. Every other agent's effective potency depends on the Observer crossing to attend to their cycle. This is the structural mechanism behind the cascade: when the Observer only holds — watching everything without ever redirecting — no cycle receives the focused attention it needs.

The four dimensions of an agent

The five functions tell you what role an agent plays. But what determines how effective an agent is at any given position? The answer comes from the cycle itself.

The cycle has four thresholds: Release, Exposure, Integration, Dissolution. At each threshold, the agent faces a binary choice — hold or cross. Did they direct energy inward or outward (Release)? Did they work with what exists or what could be (Exposure)? Did they apply universal rules or situation-specific judgment (Integration)? Did they close down options or keep them open (Dissolution)?

Four thresholds, each binary. That gives 2×2×2×2 = 16 possible profiles — one per position. An agent's profile IS their pattern of hold-or-cross decisions at the four transitions. The dimensions are not borrowed from psychology. They are produced by the cycle's own structure.

Every position requires a specific combination of these four. An agent whose threshold responses match the position's requirements operates at full strength. An agent whose responses diverge loses effectiveness — not by a fixed schedule, but by the profile distance between who they are and what the position demands.

This has a striking consequence. The sixteen personality types identified by MBTI — the most widely used personality framework in the world — are the same sixteen profiles, discovered from the psychology side. MBTI found the right dimensions because it found the threshold structure of the dissipative cycle applied to human psychology. The same 2&sup4; = 16 types because the same geometry.

But MBTI treats the profile as fixed: "you are INTJ." The framework says the profile is fixed but its effectiveness is regime-dependent. The same profile that makes you strong in Construction positions makes you weak in Encounter positions. The system moved. Your profile didn't. That distance is what the framework calls profile distance — and it is the mechanism behind every "wrong person in the wrong role" problem.

10. How nested cycles sustain each other

A single dissipative system does not exist in isolation. It is sustained by agents operating across multiple depth levels simultaneously, each running their own cycle.

Consider a star. The star itself runs a cycle at depth 1 (the four regimes of stellar evolution). But inside the star, nuclear fusion runs its own cycle at depth 2 — each fusion reaction has its own potentiality (two hydrogen nuclei approaching), construction (the strong force forming a temporary state), encounter (the nuclei fuse and release energy), and conservation (the helium nucleus maintains itself). Billions of these depth-2 cycles run simultaneously inside the star, and their combined output — energy — is what sustains the depth-1 cycle.

Below that, each nuclear interaction is governed by the four forces — gravity, strong, electromagnetic, weak — each running its own cycle at depth 3. Three depth levels, nested, running simultaneously. The deeper cycles sustain the shallower ones. The shallower cycles provide the boundary conditions for the deeper ones.

In a company: the company runs a cycle (depth 1). Inside it, each team runs its own cycle (depth 2). Inside each team, each person runs their own cycle (depth 3). Inside each person, their skills, habits, and cognitive patterns run cycles (depth 4). The person's depth-4 cycles sustain their depth-3 performance, which sustains the team's depth-2 output, which sustains the company's depth-1 cycle.

The agents at each level — Sentinel, Miner, Architect, Catalyst, Observer — are not sitting idle between their assigned actions. They are continuously running their own internal cycles to maintain their readiness. The Sentinel does not guard the boundary once and stop. It runs a continuous cycle of monitoring (potentiality), preparing (construction), detecting (encounter), and holding (conservation).

Each agent's cycle is a dissipative process in its own right, running at the depth below the agent's assigned position. The nesting is not optional — it is the mechanism by which the system maintains itself. Remove one depth level and the system above it collapses.

11. The one formula

An agent has raw potency M — how strong it is when fully deployed. And it has drain D — the fraction of its capacity pulled away by competing obligations, conservation demands from other cycles, or the inherent cost of its function.

The relationship between potency and drain is not linear. A drain of 10% costs relatively little. A drain of 50% cuts effective potency by two thirds. A drain of 80% makes the agent nearly useless.

The reason is structural: drain simultaneously reduces your force and strengthens the thing you are acting against. If you are half-drained, you bring half your force — but the system you are trying to change has been reinforced by the resources that drained away from you.

This formula is the only equation in Generative Geometry. When we apply it to the encounter between technology and humanity, it will combine with the agent structure to produce specific, testable predictions about where the gaps are and what the highest-value intervention is.

Winner takes all

Here is what the binary property means. When two systems encounter each other — a salesperson meeting a client, a drug reaching a receptor, a product landing in front of a customer — the outcome at the moment of contact is binary. You either win the deal or you do not. The drug either blocks the receptor or it does not. There is no 60% win at the point of encounter. It is zero or one.

This is different from competition. Competition is what happens before encounter — it is position 7 (Testing), the internal prototype phase where multiple configurations compete with each other to determine which one will face the world. Two salespeople competing for the same territory is Testing. The salesperson sitting across from the client is Encounter. The competition decides who shows up. The encounter decides what happens when they do.

At the agent level, every encounter is binary.

This is why "winner takes all" is visible at the agent level. When two agents compete for the same position, one holds it and the other does not. The M_eff formula determines which: the agent with higher effective potency (after drain) wins the position. The loser is displaced.

But the system is not one agent on one position. The system is sixteen positions with sixty-four agent slots. The binary outcomes at the agent level combine into the analogue outcome at the system level. The product of many zero-or-one outcomes across many positions produces a spectrum — from fully covered to nearly dead. The system's health is not binary. It is the product of many binary encounters.

In digital markets, this is visible at two levels simultaneously. At the agent level: each customer either chooses Google or does not. Binary. At the system level: the accumulation of billions of binary choices produces the gradient that makes Google's effective potency compound while competitors collapse. Network effects are the mechanism by which binary agent-level outcomes create non-linear system-level drain. Each user who joins the leader increases D for every competitor by a small amount. The hyperbolic formula turns that small increase into collapse past D = 0.5.

Digital below. Analogue above. Agents operate digitally — hold or cross, win or lose, zero or one. Systems operate analogically — the product of many digital outcomes producing a gradient. Both are real. They exist at different depths.

12. What an agent cannot control

Every dissipative system interacts with at least one other dissipative system. The star interacts with its nebula, its companion stars, its galaxy. The cell interacts with its tissue, its organ, its organism. The company interacts with its market, its regulators, its ecosystem. Interaction is not optional — it is definitional. A dissipative system that does not interact with anything is not dissipating energy and is not maintaining itself. It is dead.

Interaction means intervention. The star's radiation intervenes on nearby gas clouds. The cell's signals intervene on neighbouring cells. The company's product intervenes on its customers' lives. Every living system is, by existing, intervening on the systems it interacts with.

But intervention has a boundary. There are things you can do and things that are structurally outside your reach.

The seven layers of intervention organise what is possible:

Layers 1–5 are HOLD — understanding the system. You are diagnosing.

Layers 6–7 are CROSS — changing the system. You are intervening.

Notice the depth requirements. You need depth 3 minimum for diagnosis (you must see the agents). You need depth 4 minimum for intervention (you must be inside the agent's own cycle to influence their behaviour).

The boundary at Layer 7

At Layer 7, you have done everything within your reach. You have diagnosed the system (L1–L5), identified the best move (L6), and launched the agent to execute it (L7). The agent now runs their own cycle: they signal (recognise the move), structure (prepare), encounter (execute), and conserve (sustain the change).

But here is the structural limit: you cannot control what happens inside the agent's encounter. You can set up the move. You can select the agent. You can provide resources and reduce drain. You can motivate. But the moment the agent enters their own Encounter — the moment they actually do the thing — they are operating inside their own cycle, at a depth below yours. Their encounter produces information that depends on their specific contact with reality, and you cannot determine that information from outside.

This is Layer 8 — autonomy — and it is beyond your intervention boundary. You cannot reach it. You can only observe its output and adjust in the next cycle.

This is why management cannot control execution. It is why regulation cannot control behaviour. It is why parenting cannot control outcomes. The seven layers give you everything up to the moment of the agent's own encounter. Beyond that, the agent's autonomy is structural.

External motivation and its limit

A sales bonus is a structural intervention. The salesperson has competing demands on their attention — family, other projects, fatigue, doubt. These are drains. The bonus is designed to reduce D: it makes the target position more attractive than the competing demands. If the bonus reduces D from 0.5 to 0.3, effective potency rises from 33% to 54% of the agent's raw capability. The bonus works. The salesperson is more likely to show up.

But the bonus does not change M. A salesperson with M = 0.3 and a bonus is still a 0.3 salesperson with lower drain. They show up more. They are not better when they arrive.

This is the structural difference between external motivation and internal motivation. External motivation operates at depth 3 — it adjusts the agent's parameters (D) from outside. A bonus, a deadline, a threat, a target, a performance review. All of these reduce drain or increase the cost of not-acting. They get the agent to the encounter.

Internal motivation operates at depth 4 — it changes what drives the agent from inside. Purpose, craft, genuine interest, identity. These change M, not D. They determine what the agent does when they encounter reality.

At the moment of encounter — L8, autonomy — the bonus is irrelevant. The agent's own cycle runs. Their raw potency meets the customer. The bonus got them to the door. It cannot determine what happens when the door opens. This is why organisations that run on external motivation plateau: they can reduce drain but they cannot increase potency. They get agents to show up. They cannot make agents good.

How intervention complexity scales with depth

At depth 1 (regimes), intervention is simple: orient the system toward the right regime. Four possible states, four possible moves.

At depth 2 (positions), intervention requires knowing which of sixteen positions the system occupies and which transition is next.

At depth 3 (agents), intervention requires identifying specific agents, measuring their parameters, and computing coverage. This is where intervention becomes genuinely complex.

At depth 4 (impulses), intervention requires being inside the agent's motivational structure. You are not assigning tasks. You are shaping the conditions under which the agent makes their own choices. This is the difference between programming functions (depth 2 and 3, where you can directly assign agents to positions) and motivating agents (depth 4, where you can only create conditions and the agent's own cycle determines the rest).

The deeper you go, the less direct control you have — but the higher-resolution your intervention becomes. The trade-off is structural: depth gives you precision at the cost of control.

13. World creation

How many nested layers must a system sustain to be a world?

A single cycle at depth 1 is a process. It has four regimes, it cycles, it produces something new. But it is not a world. It is an event.

A system sustaining cycles at depth 2 is a structure. It has sixteen positions, each with distinct events, maintained by the cycling at depth 1. A star is a structure. But a star is not a world. It does not have internal agents that observe themselves.

A system sustaining cycles at depth 3 is an ecology. It has agents — distinct entities with distinct functions — filling sixty-four roles. The agents interact with each other, compete for resources, specialise. A coral reef is an ecology. A company is an ecology. An immune system is an ecology. The ecology has Observer functions (positions 14–15) that monitor the whole system. It has intelligence. But it is still bounded by a single Conservation regime.

A system sustaining cycles at depth 4 is a world. It has agents who have their own internal cycles, their own motivations, their own encounters that produce information the system cannot predict.

A world is the minimum depth at which the system creates something it cannot fully control and cannot fully observe. The agents within the world have their own dissipative processes, their own generative property, their own accumulation of specificity across cycles. The world does not just contain its agents — it creates the conditions for their autonomy, because without that autonomy the world cannot produce what the creator does not already know.

14. Autonomous agents

The world's agents must be autonomous. This is not a design choice. It is a structural necessity, and it follows directly from the generative property.

Encounter produces new information (Phi > 0) only when both sides cross independently. If one side is being steered by the other, only one side is crossing. That is not encounter. That is construction — the controller building through the agent. And construction can only produce what the controller already knows in a different form. It cannot produce encounter-dependent information, because there is no independent other to encounter.

For a world to produce something genuinely new at its deepest level — information the creator did not have and could not have predicted — the agents at that level must be free to encounter on their own terms. Their encounter must be uncontrolled. If the creator could steer every agent's encounter, the world would produce only what the creator already knew, rearranged. It would oscillate rather than evolve. The generative property requires autonomy.

This also explains why the creator cannot see the agents' work from above. If the creator could fully observe the agent's encounter in real time, the observation would constrain the encounter. It would collapse the space of possible outcomes into the outcomes the observer expects. The agent's freedom to encounter and the creator's inability to see that encounter are not separate properties. They are the same structural boundary.

What each level cannot see

The depth structure imposes a specific constraint on visibility. An agent at depth N is a combination of four phased functions operating at depth N. To see that agent — to distinguish its four functions, measure its potency, identify its drain — you need to be at depth N. You need the resolution that depth N provides.

From depth N+1 — one level below, inside the agent's system — you do not have that resolution. You are inside one of the agent's functions. You can see what that function produces: its events, its outputs, the signals it sends. But you cannot see the other three functions, because they are operating at your level, not below you.

A cell inside a liver can observe the chemical signals it receives — the hormones, the nutrient concentrations, the oxygen levels. These are events from the organ level. The cell cannot observe the liver's structure — because the cell is inside that structure.

An employee inside a company can observe the decisions that arrive — the new strategy, the reorganisation, the budget cut. These are events from the management level. The employee cannot observe the management's agency — why this strategy, why now, what alternatives were considered — because the management function operates at the employee's depth level.

An AI system trained on human-curated data can observe the patterns in its training set — the regularities, the biases, the structure. These are events from the human level. The AI cannot observe the humans who curated the data — their intentions, their selection criteria, their blind spots — because the curation is at the AI's depth level. The AI is inside the curated structure. It can see what the structure produces. It cannot see who structured it.

In plain terms: you can see what the level above you does — the decisions, the outputs, the consequences that arrive in your world. You cannot see who made those decisions or why. The events reach you. The reasoning behind them does not. A cell can detect the hormones flowing through it but not the organ that sent them. An employee can see the new strategy but not the management discussion that produced it. An AI can observe patterns in its training data but not the humans who chose what to include. The events travel downward. The agency stays at its own level.

The structural reason is derivable. An agent is a combination of four phased functions operating at depth N. To distinguish those four functions — to see the agent as an agent rather than as a set of unexplained events — you need depth-N resolution. From depth N+1 (one level below, inside the agent's system), you have only depth-(N+1) resolution. You can see what the agent's functions produce, because outputs propagate downward as events. You cannot see the functions themselves, because they operate at your level, not below it. And this constraint is structurally necessary: if you could see the agents above you and predict their behaviour, your interaction with them would be construction, not encounter. The visibility constraint is what makes Phi > 0 possible between depth levels. (The full derivation is in the First Principles paper.)

The constraint is symmetric. Looking down, you can see the agents in the systems you create — because you are at their depth or above. Looking up, you cannot see the agents in the systems that created you — because they are at your depth, and you are inside them.

And there is a structural reason this must be so. If you could see the agents above you — if you could observe their four functions and predict their behaviour — then your interaction with them would not be encounter. It would be construction: you building your response to their predicted actions. The visibility constraint is what makes the encounter with the level above genuinely encounter-dependent. The unpredictability is not a flaw. It is the mechanism that produces Phi when you interact with the depth above.

15. The yin-yang from first principles

Can the yin-yang symbol — the most ancient representation of duality — be derived from Generative Geometry?

Consider two properties of any system mid-cycle. First: where the agents are. Agents concentrate in Construction (building) and Encounter (meeting) — the two active regimes. Potentiality and Conservation have few or no active agents at this moment. Second: where the mass is. Accumulated conservation mass is lowest in Potentiality (nothing has been built yet) and highest in Conservation (everything built is now held). The mass axis lags the agent axis by a quarter turn.

These two axes, offset by a quarter cycle, produce the yin-yang shape. Construction has high agents but low mass. Conservation has high mass but low agents. The dark teardrop is where agents are concentrated. The light teardrop is where mass is concentrated. The offset between them is the structure of the symbol.

If you map agent concentration around the cycle and shade the regions with active agents dark and the regions without agents light, you get two interlocking shapes. The Construction region sweeps from position 5 to position 8, with maximum density at position 7 (Testing). The Encounter region sweeps from position 9 to position 12, with maximum density at position 10 (Discovery). The two concentrations are offset by one quarter of the cycle — exactly the relationship between the two teardrops of the yin-yang.

The seed of the opposite in each half — the small dot — corresponds to the phase-locked drain. The dominant agent in Construction carries a drain into Encounter (the seed of Encounter inside Construction). The dominant agent in Encounter carries a drain into Conservation (the seed of the next phase). Each half contains the seed of its opposite because the strength that defines one phase becomes the drain that shapes the next.

But the yin-yang is contingent, not structural. It appears when a system is mid-cycle, with agents concentrated in two adjacent regimes. A system in full Conservation (agents spread across all four regimes, maintaining everything) does not look like a yin-yang. It looks like a torus — agents distributed across all positions and all roles.

The yin-yang is a snapshot of a system in motion — specifically, a system transitioning from build to encounter. It is the most dynamic moment of the cycle, which is why it appears in so many traditions as the symbol of change. But it is one state of the deeper geometry, not the geometry itself.

16. The hourglass: every agent looks up and down

Everything in Part II can be compressed into a single image.

You are an agent. You sit at a depth. Above you: the systems that created and sustain you, intervening on your behaviour through the same seven layers you use on the systems below. Below you: the systems you create and sustain, governed by your seven layers of intervention.

The structure is symmetrical. The same eight layers extend in both directions. The same operations (hold and cross) apply in both directions. The same boundary (L8, autonomy) exists in both directions.

Looking up — you are the agent

Looking down — you are the observer

The hourglass is the view from inside a single agent who is simultaneously a creator (looking down) and a creature (looking up). The two halves are the same structure viewed from opposite ends. The waist — where you sit — is the point where looking up and looking down are the same act of observation.

Every agent has two engagements:

Engage down (CROSS): Create worlds. Build deeper. Deploy agents. Provide energy. Set boundary conditions. This is the crossing direction — working toward autonomy for the systems you create.

Engage up (HOLD): Serve the system above. Perform your function in its ecology. Contribute to its health. This is the holding direction — working to stay connected to the system that sustains you.

A complete agent does both. An agent who only looks down creates without being sustained. An agent who only looks up serves without creating. The full cycle requires both — because the system above you needs your encounter-dependent information (it needs you to create and discover), and the system below you needs your conservation depth (it needs you to sustain and govern).

Summary of first principles

Before we apply Generative Geometry to our most urgent question — our relationship with technology — here is what has been established, in order:

1. Two operations. Every dissipative system runs on hold (maintain coherence) and cross (enable interaction). No third operation has been found.

2. Four states. Two binary operations produce exactly four combinations: Potentiality, Construction, Encounter, Conservation.

3. One mandatory order. Each state's output is the next state's required input. No alternative order satisfies all dependencies. The cycle closes because Conservation, fully expressed, produces the next Potentiality.

4. Four transitions. Release (Potentiality to Construction), Exposure (Construction to Encounter), Integration (Encounter to Conservation), Dissolution (Conservation to Potentiality). Irreversible thresholds where energy is dissipated.

5. Generative property. Each cycle produces something that was not there before (Phi > 0). Encounter is the largest contributor because it combines two independent information sources.

6. Four depth levels. Regimes (4) to Positions (16) to Agents (64) to Impulses (256). Branching ratio of four at every level. No intermediate resolutions exist.

7. Intelligence born at position 14. The Observer function emerges when conservation is complex enough to require self-monitoring. Intelligence is produced by the cycle, not added to it.

8. Five agent functions. Sentinel (hold/same), Miner (hold/deeper), Architect (cross/same), Catalyst (cross/deeper), Observer (surveillance across all). Each has a structural drain.

9. One formula. M_eff = M x (1-D)/(1+D). Hyperbolic. Drain reduces you and strengthens the opponent. Winner-takes-most in coupled systems.

10. Nested cycles. Deeper cycles sustain shallower ones. Shallower cycles provide boundary conditions for deeper ones. Disruption propagates upward. Constraint propagates downward.

11. Seven layers of intervention. L1-L5 = HOLD (diagnose). L6-L7 = CROSS (act). L8 = autonomy: beyond your reach.

12. World creation. A system sustaining four nested depth levels, where the deepest agents are autonomous. The minimum structure for a world.

13. Visibility constraint. Events cross depth boundaries. Agency does not. You can see the events of the level above you. You cannot see its agents.

These thirteen principles are derived from a single starting observation: that dissipative systems hold and cross. Everything above follows logically. If the starting observation is wrong — if a third operation exists — everything collapses.

With this foundation, we can now ask the question that matters: what is technology, structurally? And what is our relationship with it?