Meet Joe: Bringing Dominus Down to One Human Life

Dominus began at the scale of nations.

The player looks across a living world and makes decisions about economies, governments, trade, infrastructure, military power, diplomacy, and the long arc of history. That scale is the point of the game. Dominus is supposed to feel like a world moving under its own weight—not a collection of disconnected menus waiting for the player to click them.

But that raised a question that eventually became one of the most ambitious features we have planned:

What would this world feel like from the perspective of one ordinary person living inside it?

Not a symbolic citizen. Not a portrait attached to a statistic. A real individual in the simulation: someone with a home, a schedule, a personality, a family, a bank balance, a job, ambitions, disappointments, memories, beliefs, and a finite amount of time.

We started calling that person Joe.

Joe could work at a gas station. He could attend school, become a mechanic, open a restaurant, run a corporation, join a community organization, get married, fall into debt, move to another city, enter politics, or—in extraordinary circumstances—become the leader of a country. The player could make Joe's important decisions, let a deterministic simulation control him, or connect an AI model and allow it to reason from the same facts and choices available to everyone else.

The more we discussed it, the more obvious it became that Joe could not be a small side mode. If we did it correctly, Joe would become the human-scale expression of the entire Dominus simulation.

The idea: place a life inside the world

The initial vision was simple: let a player create or adopt a character and watch that character live inside Dominus.

Joe would be optional. A player could engage entirely with countries and never create one. But for players who wanted the personal experience, Joe would offer something closer to a persistent virtual person: a life that continues while the player is away, reacts to the world, and develops over years of game time.

The most important clarification came later:

Players do not create a world around their Joe. They place a Joe into the one shared Dominus world and discover what that world—and their choices—turn the Joe into.

Dominus is not being designed as a collection of private worlds. There is one authoritative world, one history, one clock, one economy, one political landscape, and one population. We may eventually have thousands of player-connected Joes, but they all occupy that same reality.

That distinction changes nearly every design decision.

If two Joes apply for the same job, only one may get it. If housing is scarce, a newly created character cannot simply be handed a perfect apartment. If a player wants a wealthy executive, the system cannot invent a corporation, money, credentials, and social standing just because they were written into a biography. Those things must already exist or be acquired through the world.

A Joe can begin through several legitimate paths:

A player can adopt an existing simulated resident.
The system can materialize an individual from an anonymous population cohort.
A new arrival can enter through a conserved migration path.
A child can be born or placed into a valid household and grow up inside the simulation.

Character creation selects predispositions and world-valid starting conditions. It does not guarantee outcomes.

That may sound restrictive, but it is actually what gives the feature meaning. Becoming a business owner matters because the business had to be built. Winning an election matters because voters, institutions, money, time, opponents, and law all existed before the campaign. A relationship matters because it developed between two persistent people rather than appearing as a decorative status.

From gas-station clerk to head of state

The range of possible lives is intentionally broad.

A Joe should be able to:

attend school and earn evidence-backed qualifications;
apply for work, accept an offer, attend shifts, earn wages, quit, or be dismissed;
rent or own a home and move when circumstances allow;
buy goods, borrow money, repay debt, save, invest, and experience financial pressure;
build friendships, join a household, marry, separate, raise children, care for relatives, age, become ill, recover, and die;
start and operate organizations or businesses;
participate in civic, religious, professional, political, or illicit organizations;
enter public service, law enforcement, activism, crime, corporate leadership, or politics;
campaign for office, win an election, serve a term, lose authority, or rise as far as the country's institutions permit.

The political possibility was especially interesting. In a world with roughly two hundred countries, the eventual heads of government could be controlled by different players, deterministic policies, or AI controllers. A Joe who legitimately becomes president, prime minister, monarch, governor, mayor, or another officeholder receives the same category of country controls available to an equivalent country player—but only through the authority of that office.

That authority is never permanent or magical. Constitutions, elections, term limits, legislatures, courts, parties, public support, budgets, information, and succession rules still apply. A four-year term remains four game years. Losing office revokes the powers that came with it.

This gives Dominus a remarkable bridge between scales. A national policy changes the lives of individuals; individuals can organize, vote, build institutions, and eventually change national policy.

Why we rejected a giant list of hard-coded commands

Early on, it was easy to imagine Joe as a list of commands:

Apply for job
Accept job
Quit job
Enroll in school
Attend work
Move residence
Buy
Borrow
Start a business
Propose a relationship
Marry
Separate

Those actions are useful, but a fixed command list becomes a trap. The moment Joe opens a restaurant, the player needs restaurant decisions. A car wash needs different decisions. A hospital, trucking company, church, political party, criminal enterprise, and software firm all introduce different operations.

Hard-coding every possible life would produce an enormous switch statement that could never become truly expandable.

The better design is a layered capability system.

At the bottom are universal primitives: people, roles, organizations, locations, schedules, money, inventory, contracts, authority, relationships, communication, movement, and events. On top of those primitives, versioned capability modules define what a particular job, institution, or organization can do.

A restaurant and car wash can therefore share the same accounting, employment, scheduling, property, purchasing, and customer-demand systems while still exposing different operational choices.

A restaurant may manage menus, ingredients, food preparation, seating, spoilage, inspections, and meal-service capacity. A car wash may manage water, chemicals, equipment uptime, throughput, vehicle queues, maintenance, and environmental rules. They are not separate games. They are different capability packages built from shared rules.

The same pattern applies beyond business. A police department, university, political party, neighborhood association, religious community, charity, union, or high-control organization can share membership, leadership, funding, property, communication, reputation, recruitment, discipline, and succession mechanics while adding the capabilities unique to its purpose.

This is how we aim for “limitless within reason”: not by allowing arbitrary text to become reality, but by making new content composable.

Designing a person instead of choosing a destiny

A player should have substantial freedom when creating a Joe. The system should support appearance and presentation, temperament, personality, values, worldview, aptitudes, flaws, background preferences, relationships, starting goals, and control policy.

But we do not want players choosing a destiny from a menu.

There should be no “criminal trait,” “terrorist trait,” “cult leader trait,” “good person” slider, or genetic switch for morality. A police officer is not automatically honorable. A business owner is not automatically greedy. A deeply religious person is not automatically peaceful or violent. Nationality, ethnicity, religion, sex, gender, disability, and culture cannot be shortcuts for intelligence, loyalty, morality, or political behavior.

Who a person becomes should emerge from several interacting layers.

Personality

The current design includes bounded dimensions such as:

openness;
conscientiousness;
sociability;
agreeableness;
emotional stability;
risk tolerance;
ambition;
patience;
rule adherence;
empathy;
dominance;
impulsivity.

These dimensions influence preferences, not permissions. High dominance does not make someone a capable leader. Low empathy does not automatically cause violence. High rule adherence does not guarantee moral behavior, especially under an unjust system. Impulsivity does not let a character bypass consequences.

Values

Values express what a person tends to care about: family, wealth, security, status, power, autonomy, community, equality, career, education, leisure, public service, tradition, religious or philosophical purpose, national identity, freedom, and achievement.

Values can conflict. Joe may value family and career but have time for only one urgent commitment. He may value security while considering a risky opportunity. He may care about public service but also desire status. Those tensions are where interesting decisions appear.

Worldview and ideology

We considered whether ideology should be a single value. It quickly became clear that one left-versus-right number would be too crude for a global simulation.

Instead, worldview can be represented through several independent axes:

authority versus liberty;
hierarchy versus equality;
tradition versus change;
nationalism versus universalism;
collectivism versus individualism;
market allocation versus redistribution;
secular public order versus sacred or religious public order;
trust in institutions;
tolerance of disagreement and pluralism;
belief that peaceful institutional change is effective.

Separate properties describe conviction, rigidity, openness to contrary evidence, identity fusion, grievance, zero-sum thinking, and target-specific moral disengagement.

Even this is not a destiny system. Strong conviction does not create followers. Grievance does not create a target, money, authority, knowledge, or an executable harmful action. A radical trajectory requires causes, relationships, organizations, opportunities, repeated choices, and failures of protective institutions. A similar person in a different family, school, community, or political system may develop very differently.

Aptitude and competence

Ability must remain separate from personality.

Aptitudes can affect how easily a person learns reasoning, communication, organization, finance, technical work, art, physical skills, care work, or civic activity. Competence comes from education, training, credentials, practice, and experience.

Charisma may help someone persuade, but it does not decide what they believe. Intelligence may help someone plan, but it does not decide whether the plan is generous, selfish, legal, or harmful. Security knowledge, political expertise, and knowledge of illicit markets must be learned through actual experiences and access.

Needs, emotions, goals, habits, and memory

People do not make decisions from personality alone.

A Joe also has changing needs: food and basic consumption, housing security, liquidity, safety, health, rest, belonging, care responsibilities, employment security, achievement, autonomy, and relief from stress.

Short-lived emotional states—fear, anger, grief, confidence, fatigue, loneliness, satisfaction—can temporarily shift priorities, but they must have a cause and decay over time.

Goals give direction. Commitments create obligations. Habits add inertia. Memory records what the person actually experienced or learned. Trust, affection, loyalty, gratitude, and grievance are tied to specific people, organizations, and events rather than stored as one universal number.

The result is not a character sheet that predicts a life. It is a starting structure capable of producing one.

How a deterministic Joe decides

A deterministic character should not follow a hand-authored script. At each meaningful decision point, the simulation constructs the actions currently available to that person.

The process looks roughly like this:

Build an authorized observation of the Joe's situation.
Generate current affordances: the things Joe can actually attempt now.
Remove options that fail hard requirements such as time, money, law, authority, access, knowledge, consent, or physical possibility.
Estimate the likely consequences of each remaining option.
Score those consequences against needs, values, worldview, goals, relationships, obligations, habits, personality, risk, and expected future value.
Choose deterministically using stable ordering and controlled uncertainty where appropriate.
Submit the result through the same authoritative action system used by every other controller.

“Do nothing” and “defer” must be real choices. Sometimes the rational decision is to wait. Sometimes procrastination, exhaustion, fear, or conflicting obligations cause a missed opportunity.

Most importantly, the decision system can explain itself. A result might say that Joe accepted a night shift because urgent rent pressure outweighed his preference to remain in school, but that the new schedule increased fatigue and reduced study time. That explanation points to real inputs and real consequences.

One action pipeline for humans, simulation, and AI

Joe supports three main control modes:

Human control: the player chooses important actions.
Deterministic control: the simulation chooses from Joe's current profile and circumstances.
AI control: a connected language model reasons from the same authorized observation and affordances.

There may also be standing policies: instructions such as prioritizing school, avoiding debt above a limit, accepting emergency medical care, or never making a major relationship decision automatically.

All controllers use the same action envelope. None of them can mutate the world directly.

The central chain is:

observation → available choices → proposed action → validation
→ reservation of costs and capacity → authoritative receipt
→ state mutation → downstream reaction

If an AI model says Joe should buy a building, the server still verifies that the building exists, is for sale, that Joe knows about it, can afford it, has authority to act, and has enough time to complete the transaction. The model cannot invent money, facts, consent, followers, or outcomes.

AI control is also planned as bring-your-own-provider. A player who wants an AI-controlled Joe connects and funds an approved model account. OpenRouter-style account authorization is the preferred experience, with direct provider keys treated as an advanced option. Credentials must remain server-side, encrypted, revocable, budgeted, and isolated between players.

If the model is unavailable, times out, returns malformed output, reaches its budget, or is disconnected, Joe falls back to deterministic control. One player's provider can never freeze the shared world.

Every action needs a reaction

Joe inherits the central simulation law of Dominus: every action has a reaction.

A job is not merely a label. It occupies time, consumes commute capacity, produces labor, creates wages, affects business output, incurs taxes, and changes future opportunities.

School consumes time and institutional capacity, may cost money, produces learning and credentials, and affects employment prospects.

Opening a restaurant consumes capital, property, equipment, utilities, ingredients, labor, permits, and management attention. Its success or failure affects workers, suppliers, customers, tax revenue, competitors, and the local economy.

A political campaign consumes money, time, organizational capacity, attention, and public trust. Winning transfers defined authority for a defined term. Policies then flow back into households, firms, schools, prices, public services, and future elections.

Technically, every accepted action produces a receipt recording its authoritative inputs, cost, reservation, before-and-after values, immediate effects, downstream events, and visibility. Rejected actions produce explainable receipts as well.

This receipt system is essential for debugging, replay, player trust, and the Life Feed. It is how the game can answer not only “what happened?” but “why did it happen?”

Reconciling one person with billions

A global life simulation cannot represent every human at maximum detail. Dominus therefore needs two compatible resolutions.

Most of the world's population remains in anonymous aggregates: age cohorts, labor pools, households, consumers, voters, and other statistical groups. Adopted, controlled, narratively important, or institutionally important people become sparse explicit entities.

When a Joe becomes explicit, the simulation must carve that person out of the aggregate exactly once.

authoritative total = anonymous remainder + explicit allocations

The invariant applies to population, labor, income, consumption, housing, votes, education, service demand, production, taxes, and other material dimensions. If Joe earns a wage explicitly, the macro economy cannot also count that wage as though he remained anonymous. If he occupies a home, that capacity cannot remain available to somebody else.

The same applies in reverse when a person migrates, dies, or is safely dematerialized. Assets, liabilities, contracts, relationships, roles, and population shares must reconcile without duplication or disappearance.

This is one of the least visible parts of Joe and one of the most important. Without it, the personal simulation would be a decorative layer floating above the real game. With it, Joe becomes part of the economy.

Time: one real day, one game week

Individual lives forced us to reconsider game time.

The current target is:

1 real day = 7 game days
server downtime or intentional pause = zero elapsed game time

At that rate, a player who checks once per real day returns after approximately one week of Joe's life. That is enough time for work shifts, school, household events, purchases, conversations, and new opportunities without making an entire life disappear overnight.

The clock is planned at game-minute resolution even though larger systems may settle daily, weekly, monthly, or yearly. Joe can attend a shift, commute, sleep, study, and meet someone on sub-day schedules while the macro economy runs at an appropriate cadence.

The two ideas—pace and resolution—are intentionally separate. The world may advance seven days per real day while still processing personal events in fifteen- or sixty-minute game intervals.

If the server is down, the world pauses. There is no wall-clock catch-up on restart. Nobody loses a job, misses an election, ages a year, defaults on a loan, or dies because the service was unavailable.

Major player decisions should usually remain open long enough for a daily visitor to respond. Routine obligations continue automatically. Emergencies use standing policies or deterministic fallback rather than stopping the global clock for one person.

The Life Feed

The Life Feed is where all this complexity becomes personal.

Instead of presenting Joe as a wall of statistics, the player sees a chronological stream of moments:

Joe completed his first week at a new job.
Rent was paid, leaving less money than expected.
A teacher recommended a different program.
A friend introduced Joe to an organization.
A relationship changed after a specific event.
A business opportunity appeared.
An election result changed Joe's job or neighborhood.
A major decision is waiting.

Some entries summarize routine activity. Others become decision cards with a clear choice. A “Next Moment” button can reveal the next already-resolved event or current decision, but it can never privately advance Joe or invent an outcome. There is only one world clock.

The feed is a projection of authoritative history. Every important card links back to receipts and events. Text can explain and dramatize what happened, but presentation never becomes the source of truth.

Generated images and video can make the feed feel alive, provided they preserve character identity, respect privacy, pass moderation, and remain labeled presentation. A generated picture of Joe's graduation illustrates an event that already happened; it cannot prove the graduation or alter the credential.

The long-term vision includes a dedicated Joe Feed application: a simpler place to follow one or several lives, receive decision notifications, view weekly recaps, inspect important consequences, and interact with the same characters that exist in the main Dominus world. It would be another client for the same backend—not another simulation.

Thousands of Joes without thousands of constantly running minds

A shared world with thousands of explicit people creates a serious scale problem if every character is evaluated continuously.

The answer is event-driven simulation.

A Joe can operate at several levels of attention:

Background: routine schedules, contracts, bills, and obligations execute deterministically or through aggregate settlement.
Active: a material event wakes the person for a detailed decision.
Focused: a player is watching, a major choice is pending, a conversation is occurring, or an AI controller is needed.

The scheduler wakes only people and systems with something due. It does not scan every person every game minute, and it does not call an AI model for routine behavior.

Internal processing may be partitioned by region for performance, but that is invisible infrastructure. People retain global identity. Migration, commerce, elections, relationships, and history continue across regional boundaries.

Our benchmark targets are deliberately larger than the initial audience: large numbers of read-only citizen lives, thousands of active deterministic Joes, bursts of simultaneous personal actions, hundreds of live feed connections, provider outages, restore and replay, and long-running worlds with bounded storage.

Scale cannot be purchased by weakening causality. If the system falls behind, it must expose lag and backpressure rather than silently skipping consequences.

The deeper vision

Joe began as a way to see the game from street level. It has grown into a framework for connecting the smallest and largest parts of Dominus.

A recession is no longer only a chart. It is Joe losing hours, reconsidering school, delaying a move, taking on debt, or joining a political movement.

A new transit system is not only an infrastructure score. It changes commute time, job access, business traffic, housing demand, family schedules, and local opportunity.

A war is not only a national event. It affects safety, prices, migration, public opinion, employment, grief, institutions, and the choices available to individuals—without turning Dominus into a tactical battle simulator.

A political reform can alter whether a frustrated person believes peaceful change is possible. A mentor can redirect a life. A corrupt institution can erode trust. A supportive household can absorb a shock. A persuasive organizer can build a movement, but only by finding people, earning influence, obtaining resources, and operating within—or against—the institutions around them.

This is why we do not want a “become cult leader” button or a “become president” preset. The interesting part is the path.

The player designs a person, chooses when to intervene, and watches the interaction between personality, circumstance, institutions, relationships, and accumulated decisions. Some Joes will thrive. Some will fail. Some will surprise their players. Some will become admirable, compromised, dangerous, ordinary, powerful, loved, isolated, or forgotten.

And because they all inhabit the same world, their lives can collide.

That is the promise of Joe: not a story pasted onto Dominus, but a human life generated by the same world that moves nations.