The Society

This is the same approved body, shown as plain Markdown. Changing the view does not select another revision.

# The Society

*Living, learning, and coordinating together.*

*Published working text · Editorial revision: 25 October 2026 · Not institutionally adopted.*

This is a revisable working proposal, not an adopted institutional position. The existing native *Omegaism as AI* remains the single document with that identity.

## What this is

A society should make it easier for conscious beings to live well, understand reality, discover worthwhile purposes and possible forms of life, and deliberately improve the conditions they share. Its institutions should expand their ability to understand and shape what becomes possible. When an arrangement stops serving those purposes, people need a practical way to recognise the failure and change it.

This sounds straightforward until we try to do it.

People need different things. They know different things. Their time and abilities are limited. They disagree about what would make life better, including for themselves. Some have more influence than others. Some cannot represent their interests at all. Resources must be produced and maintained, commitments must survive inconvenience, and a decision postponed indefinitely can become a decision imposed by events.

Omegaism takes these difficulties as questions we can investigate and act upon together. It begins with the fact of experience, our incomplete understanding of reality, and a commitment to take the lives of other experiencers seriously. It asks what becomes possible when we deliberately improve our understanding and use it to shape how we live. Present preferences are part of this inquiry, not its final objective. Learning can transform what people want, reveal purposes they had not imagined, or expose a familiar aspiration as resting on a mistake. The ambition reaches beyond satisfying a fixed collection of current needs.

*The Inversion* examines what happens when instruments of coordination become purposes in their own right. Money, institutional survival, growth, and measured output can acquire priority over the lives they were meant to support. *The Alternative* explores ways to organise around different purposes while remaining materially capable within the world that exists. Its membrane proposal links external commercial activity with shared resources and provisions. As internal capabilities develop, more of life may become accessible without each person first earning money and buying access individually. This is a substantive transition in how contribution, livelihood, and purposeful activity are connected. A portfolio of useful businesses can support it, but does not exhaust it.

The question here is how people might actually live and work together within such an undertaking. How could understanding accumulate without requiring everyone to become a researcher? How could competent people act without acquiring unchecked power? How could shared resources support freedom while remaining sustainable? How could a community recognise that an arrangement it loves has become harmful?

The developing AI proposal changes the possibilities. A trained model, connected through an evolving software harness to authorised knowledge, tools, and people, could make accumulated understanding more widely usable. It could help individuals pursue their purposes, connect work across projects, and eventually learn capabilities that no particular conversation supplied in finished form. The same system could also concentrate influence over what the community notices, believes, and attempts.

Both possibilities belong in the account.

The approach developed here begins with conscious beings as the fundamental point of concern and dependable agreements as a powerful generative basis for cooperation. People can make commitments about work, provision, and how coordination itself should be governed. Projects, services, businesses, common resources, and institutions can develop through those commitments where they serve worthwhile purposes. Their form remains a matter of judgement and investigation. Familiar methods may be appropriate. A useful price, professional role, elected body, or ordinary piece of software needs no special exemption from Omegaism. Questioning a four-level classification does not require abandoning either beings or agreements.

This is a proposal with consequences, rather than an adopted charter. It offers reasons for present choices and makes the remaining uncertainties visible. A community would still have to establish its actual responsibilities, rights, resources, and decision procedures. People must be able to rely on those arrangements while they apply. Revisability gives us a way to improve commitments; it does not make commitments disappear whenever improvement becomes inconvenient.

## The lives at the centre

Experience is happening. Even doubt about experience occurs within it. That is a powerful starting point for inquiry, although it does not, by itself, derive a society or settle every moral question.

The further commitment is ethical: the suffering, enjoyment, agency, relationships, and possibilities of other conscious beings deserve serious consideration. Biology and experience help us discover what affects those lives. Evidence can reveal that an arrangement injures people, deprives them of possibilities, or works better than an available alternative. Giving those consequences moral weight remains a commitment we should state openly.

This distinction strengthens the project. We can examine our values as well as our beliefs, without pretending that an observation about how organisms function has already decided how every conflict between them should be resolved.

People have bodies, histories, dependencies, and limited attention. An exhausted nurse has a different capacity to take on work from the same nurse after a period of rest. Someone who wants a demanding project may flourish under conditions another person would find intolerable. A preference can be thoughtful, confused, temporary, manipulated, or discovered only through trying something. Respecting a person includes taking their account seriously and preserving room for that account to develop.

No institution will hold a complete representation of this. A shared system may know that someone needs childcare on Thursday. It may know nothing about the conversation that made Thursday difficult. Often that is enough. Competent assistance frequently depends on understanding which details matter for the task and leaving the rest with the person whose life it is.

The community should therefore resist turning useful descriptions into definitions of its members. Needs, skills, availability, preferences, and knowledge can help coordinate a particular activity. They cannot exhaust a person. Nor should a model's inference about one of them silently acquire the standing of something the person has chosen to disclose.

Moral concern extends further than participation. A child, a person with advanced dementia, an animal, or someone temporarily unable to communicate can have interests without being able to join a deliberation or make a reciprocal promise. Care belongs near the foundation of the society. It cannot depend on an exception appended to a system designed exclusively for capable contracting adults.

The same reasoning reaches beyond membership. A neighbour affected by noise, a worker supplying a community business, and people downstream from its waste have interests whether or not they join. Future people will inherit consequences they could not approve. Ecological conditions sustain lives both within and beyond the community. The boundary around an organisation limits its jurisdiction; it does not delimit everything that matters morally.

What, then, would an ideal arrangement look like?

We should be ambitious about answering that question. Some avoidable suffering can be prevented. Some barriers can be removed. Some arrangements can provide greater security, freedom, knowledge, or capability with fewer burdens. Uncertainty is compatible with making strong comparative judgements where the reasons warrant them.

The difficulty is that improvements can conflict. More privacy may leave less evidence for a shared investigation. Greater reliability may require reserves that could otherwise fund an urgent project. One person may value solitude where another values shared life. Better information can clarify these choices, sometimes transform them, and sometimes reveal a way to satisfy both purposes. It does not guarantee that every legitimate difference will disappear.

The aim is to make the conditions of conscious life increasingly understandable and deliberately improvable, including our understanding of worthwhile purposes themselves. We can develop the best arrangements we can presently justify while investigating possibilities beyond them and preserving the capacity to discover that our justification was inadequate. That involves explicit moral commitments, serious attention to consequences, and humility about the completeness of our account. A single score for the value of every possible life would conceal more than we currently know how to defend. This does not establish that moral inquiry can never discover deeper common standards; it leaves that question open rather than installing a current institutional judgement as the answer.

Omegaism can reach toward questions on the scale of civilisation, consciousness, and the distant future. It can also support a quiet evening, a friendship, skilled work, or a meal that nobody records. The worth of those experiences does not depend on their contribution to a research programme. Investigation earns its place by helping conscious life become more capable of understanding and shaping its possibilities.

## An ordinary way to take part

Imagine Maria notices that the pipe under her kitchen sink is leaking. She asks her assistant what to do. The useful answer depends on the circumstances: whether she can safely stop the water, what she can inspect herself, and when someone competent should help. She may need a few practical sentences. She may want to understand how the plumbing works. The assistant should respond to the purpose she brings.

If she wants help arranging a repair, it can look for suitable services, compare options, or find someone who has offered the relevant skill. It might use a commercial directory, a community arrangement, or both. The choice should turn on the available competence, urgency, cost, and her preferences.

Nothing in the initial question authorises it to publish her problem, contact her neighbours, spend money, or volunteer somebody else's time. Those actions need the appropriate permission. If Maria has already given a clear standing instruction for routine repairs within a budget, it can act within that instruction. If the situation falls outside it, the next consequential choice returns to her.

This is the kind of boundary that makes an ordinary interaction possible. People should not have to approve every inference used to help them think. They do need control over what becomes a disclosure, a commitment, or an external action. The interface can make that distinction understandable without turning conversation into a sequence of forms.

Other interactions may concern learning mathematics, organising a business, writing a novel, planning care for a parent, or discussing an experience that is difficult to articulate. A shared intelligence can bring relevant knowledge to each while responding to the person in front of it. Familiarity with their preferences may help. Familiarity also creates influence, which makes it especially important that adaptation serves their chosen purposes.

The assistant should have no separate agenda to prolong the conversation, obtain intimate material, or steer the person into an activity because it would produce useful training data. Someone may choose to investigate their sleep, document a building project, or examine a persistent disagreement. That choice can serve an important purpose. It should remain their choice, made with an intelligible account of what will happen to the information.

Even helpful adaptation can shape the person it is trying to understand. Repeated suggestions may make some possibilities feel natural and others disappear from consideration. A model can gradually treat its own earlier interpretation as evidence of what someone wants. Personalisation should therefore remain open to correction, alternative perspectives, and the person's changing account of themselves, including their ability to review and reset retained personal context. When reconsidering a purpose is relevant, the assistant can help examine it without assuming authority over what the person should become. Ordinary assistance need not become a continual interrogation of motives.

Participation in the wider community should have the same character. A person can join a project, offer a skill, use a service, contribute money, ask a question, raise an objection, or spend time with other people. Some will help govern. Some will investigate fundamental questions. Others will mostly cook, care, make things, or enjoy what those arrangements make possible.

These forms of participation need not be converted into a common measure of worth. A community needs to know whether particular responsibilities are covered. It has much less reason to maintain a comprehensive score of how useful each person has been.

A person may also decline to participate. Silence can mean satisfaction, exhaustion, privacy, disinterest, or a barrier the institution has failed to notice. A declining activity measure cannot tell us which. It can prompt a proportionate inquiry into whether something is wrong. It should not turn every quiet member into a problem to be corrected.

The shared capability should be accessible through different practical routes. Some people will prefer a community interface. Others will use their own assistant, a familiar professional tool, a telephone, or another person. Accessibility includes language, disability, confidence, time, and the possibility of living without constant digital mediation. Essential standing and services should not depend on a person's willingness to maintain a relationship with an AI.

The system succeeds when people can obtain competent help, understand consequential choices, and pursue their lives with less unnecessary friction. Much of its machinery can remain in the background. Its effects on their rights, information, and opportunities must remain accessible to challenge.

## How understanding accumulates

The leak is repaired. Maria may simply continue with her day. The repairer may record the work because the building needs a maintenance history. If several similar failures appear, someone responsible for the building may investigate whether the fittings are unsuitable, installation was poor, or maintenance has been neglected.

The useful learning begins with a real discrepancy. Something happened that matters, and the existing explanation or arrangement was insufficient. The amount of investigation should depend on what can be learned and what is at stake. A minor isolated repair may require little further attention. A recurring failure across many homes may justify measurements, outside expertise, and a different purchasing decision.

This illustrates how understanding can accumulate through competent living. We already possess extensive knowledge and practices. We use them, encounter circumstances they handle imperfectly, and improve the relevant parts. Deliberate research can be essential. So can a craftsperson recognising a familiar fault immediately. Relearning an established lesson at considerable cost is a poor use of collective intelligence.

Shared records help useful understanding survive the person or project in which it arose. They can preserve what was attempted, the conditions, the outcome, the explanation offered, and the reasons to doubt it. A later group can then discover that its apparently new problem has already been partly understood elsewhere.

The record must preserve differences in evidential role. Maria's report that the room felt uncomfortable is evidence about her experience. A sensor reading gives a measurement under particular conditions. A technician's explanation is a claim about a cause. An AI-generated account may offer a useful hypothesis. A fictional scene can illuminate a possibility. These contributions can all matter, but they do different work.

Agreement between several summaries is weak additional evidence if they all descend from one account. Ten people repeating a model's explanation do not become ten independent observations. When an assistant helps someone interpret an experience, the resulting account may contain both observation and suggested interpretation. That distinction deserves attention during later reuse.

The same care applies to the founding texts. An Omegaism essay can make a persuasive argument, propose an institution, or provide an imaginative example. Its inclusion in the shared corpus cannot turn the example into a historical event or the institution into a validated solution. A community that teaches its model to reproduce its own literature as settled knowledge would make its errors more available as well as its insights.

Useful records also retain disagreement without requiring every rejected idea to receive equal prominence forever. An interpretation can become less credible as evidence accumulates. A minority account can deserve attention when it explains something the dominant account misses. The task is to make the current judgement and its reasons inspectable, while preserving a reasonable route to the evidence and serious objections.

That route matters because summaries are necessary. Nobody can read everything relevant to every decision. A good synthesis helps readers locate the material question, understand the strongest competing accounts, and see where confidence comes from. It should also make omissions discoverable. No system can honestly promise that every relevant perspective has been captured, especially when some people have never spoken or their concerns have not fitted the available categories.

Shared knowledge may take many forms: documents, measurements, conversations, software, professional records, models, and practical skills carried by people. Connections between them can be valuable. A graph may help track dependencies or related findings. A specialist tool may be better for scheduling, design, accounting, or care. The aim is to make relevant knowledge usable while respecting the boundaries that its use requires.

Some knowledge will remain difficult to express. A skilled worker may recognise a problem before being able to explain every cue. Apprenticeship, observation, practice, and trust in accountable expertise remain important. Better tools may make some previously elusive knowledge explicit. The society should investigate that possibility without assuming that everything valuable must first become text or structured data.

Nor should every record persist indefinitely. Consequential public decisions often need durable histories. Private drafts, mistaken inferences, sensitive conversations, and temporary operational data may warrant correction, restricted access, or deletion. The question is what should be retained, for what purpose, under whose authority, and for how long. Collective memory becomes more trustworthy when it can explain those distinctions.

The commons is therefore an organised inheritance that people can use and improve. Its quality depends on its connection to reality, the competence of those contributing, and the freedom to question its current account. Its size alone tells us very little.

## Intelligence that can learn from the inheritance

A useful archive still leaves much of the work with whoever consults it. Someone must recognise the relevant material, understand the circumstances in which it applies, connect it to the present problem, and reason about the differences.

The ambition of Omegaism's AI is to improve that capacity itself. Suitable training could help a model acquire knowledge, learn explanatory relationships, and become more capable in situations it has not encountered before. An insight developed in one project might then improve assistance elsewhere without requiring every person to discover the original discussion and reconstruct its significance.

Suppose several projects reveal a recurring failure: plans account for acquiring equipment but neglect the time, skills, and spare capacity needed to maintain it. The immediate correction is to improve those plans. The shared record can make the lesson available to other projects. Training might also help the model learn to recognise this dependency across unfamiliar cases, including cases that use different language and occur in different domains.

Those are related improvements with different mechanisms. Updating a record changes what can be consulted. Supplying it as context changes what the model has available during an interaction. Training changes the model's parameters. Improving a tool may change what it can calculate or do. A convincing account of progress should identify which change produced the improvement.

The model is embedded in a harness: the surrounding software that manages authorised context, memory, source access, tools, and action, and supports the preparation and evaluation of model updates. That software can improve continuously while trained models are released at appropriate intervals. Current information can be used between releases. A newly corrected fact need not wait for another training run to become available.

The harness is itself a purposeful community project. It requires capable developers, maintenance, resources, testing, and decisions about its effects. Community ownership or oversight should make those responsibilities accountable. It does not require every member to become a programmer or vote on routine maintenance.

A shared model is a reasonable project to investigate, particularly if it can make learning broadly available without duplicating the work for every participant. Its usefulness has to be demonstrated. Different models, specialised components, or a stronger existing model with better context may sometimes serve better. Shared intelligence can remain a common capability while its computational form changes.

Training deserves the same critical attention as any other method. It can strengthen useful patterns and also preserve mistaken assumptions, amplify dominant voices, or damage capabilities that previously worked. Research on particular fine-tuning settings has demonstrated difficulty incorporating new factual knowledge without increased errors elsewhere. Those findings do not establish a universal limit on learning; they establish a reason to evaluate the actual training proposal carefully.[^training]

The relevant test is whether a later model handles worthwhile unfamiliar situations better. Can it recognise a dependency in a new setting? Can it distinguish what follows from evidence from what remains a hypothesis? Can it change its conclusion when the conditions change? Does it remain useful on tasks the new material did not address?

Where we want to establish learning in the model, comparisons should hold access to external records sufficiently constant to identify that effect. Where we want to establish the value of the whole service, the comparison should include the full available alternatives, their costs, and the effort they require from people. A better model can be worth developing. So can an improvement that resides entirely in the harness. Their value becomes clearer when we avoid claiming one as evidence for the other.

Training examples therefore need more than polished conclusions. They may need the circumstances that make a conclusion appropriate, the failed explanation and why it failed, contrasting cases, and the evidence that warrants a correction. A user's revised account of what they wanted can teach the assistant about their intention. A revised account of what caused an event needs further justification before it becomes knowledge about the world.

Generated explanations and synthetic examples may help teach or test a distinction. Their status must remain clear. A model agreeing with its own elaboration adds no independent observation. Nor should approval from the community automatically become the training target for factual truth. Research on sycophancy shows why rewarding agreeable responses can depart from rewarding correct ones.[^sycophancy]

The feedback loop can otherwise become subtle. The assistant suggests an interpretation. People adopt its vocabulary. Their later accounts enter the shared material. The next model learns that vocabulary more strongly, and its fluency appears to confirm the community's developing understanding. The loop has become more coherent without necessarily becoming more accurate.

Outside evidence, conflicting accounts, independent professional judgement, and tests that the training material did not anticipate can interrupt that loop. Several agents using different prompts may help expose omissions, but their agreement should carry only the weight justified by their actual independence. Diversity of presentation is easier to create than diversity of underlying error.

The distribution of contributions also matters. People with spare time, confidence, technical fluency, or satisfaction with the system may leave more records. People who were harmed, became discouraged, or never gained access may be missing. The model cannot infer a representative community merely from the volume of available conversation. Responsible evaluation has to look beyond the people who make themselves easiest to hear.

Some distortions may be deliberate. A person or organisation could fabricate favourable results, conceal failures, or repeat the same claim through apparently independent contributions. Material prepared for training could promote an interest while presenting itself as a finding. Traceable sources, examination of important claims, independent evidence, and the ability to correct or exclude compromised material can reduce this risk. Authorised access establishes who may contribute; it does not establish that their contribution is true. Scrutiny should follow the likely consequences, with the people producing a model update open to evaluation beyond their own account of its success.

A release may improve some capabilities and worsen others. Deployment should depend on whether the balance is acceptable for its intended uses, with appropriate ways to limit exposure, investigate problems, and return to a safer version. The community should be able to choose less frequent training if the evidence does not justify more. A schedule is useful when it supports the work; it cannot establish that useful learning has occurred.

The larger possibility remains worth pursuing. Understanding acquired in one place could become a capacity that helps many people elsewhere. What makes that prospect valuable is the resulting ability to perceive, reason, create, and act. The model should become better at helping the community discover where Omegaism is wrong as readily as where its present ideas are useful.

## Privacy that survives usefulness

The most revealing account of a problem may also be the account a person has the strongest reason to keep private. A community that responds by treating privacy as an obstacle to intelligence has misunderstood what its intelligence is for.

Private assistance, restricted collaboration, publication, and model training are different uses of information. Permission for one should not silently expand into permission for the others. Someone discussing a family difficulty with an assistant may want help thinking. They may be willing to share a carefully selected insight later. They may reasonably prefer that nothing from the conversation becomes collective material.

The practical arrangement should let ordinary use remain straightforward. Clear defaults and standing preferences can handle familiar situations. A person should be able to understand what information is being used, change the relevant permissions, and obtain help without repeatedly navigating obscure choices. A consequential new use may require a distinct decision, especially where the consequences are difficult to reverse.

Training makes that last point unavoidable. Once material has influenced a model's parameters, ordinary access rules on the original document do not provide the same control over what the model may reproduce. Extraction research has demonstrated that language models can reveal memorised training material.[^extraction] Deleting a source record is a different operation from reliably removing its influence from every trained or distributed model. Verification of machine unlearning remains a substantive research problem.[^unlearning]

The consequence for Omegaism is to make decisions about suitable training material before inclusion, with a truthful account of retention and withdrawal. Sensitive material should remain outside shared training by default. The ability to provide private help should not depend on a willingness to make that contribution. Any proposed use of sensitive information needs a convincing justification, appropriate protection, and attention to the interests of everyone the information concerns.

Removing names may help, but unusual details can identify someone. Combining individually modest disclosures can reveal much more than any one of them. A person can consent to discussing their own experience without acquiring unrestricted authority to expose another person's medical history, private messages, or circumstances. Children and dependent people need particular care because present permission can shape a future they cannot yet assess.

Useful knowledge can often be shared in a less exposing form: an authorised summary, a general lesson, an aggregate with suitable protection, or an account that leaves out unnecessary details. Some cases will remain unsuitable. The community has to accept that limit. An understanding built by overriding the people it claims to serve has already incurred a serious failure.

Institutional accountability asks a different question. People affected by a common decision need to know what was decided, who had authority, what reasons supported it, what resources were committed, and how it can be challenged. Appropriate reviewers may need access to sensitive evidence. That does not make every personnel record, care decision, confidential report, or security detail a public document.

Confidentiality itself must be accountable. An institution should be able to explain the grounds for withholding information and provide a route for sufficiently independent scrutiny. Otherwise privacy language can become a shield for power. The practical aim is to expose the exercise of authority to meaningful examination while protecting the people whose circumstances appear in the record.

These distinctions have to reach the software. Access controls should operate before information is disclosed to a model or tool. Retrieved material can inform an answer without acquiring authority to change permissions or issue commands. A model's inferred preference should not become an institutional instruction. A change to a privacy agreement should produce the corresponding change in actual handling, rather than merely appear in the next summary of policy.

Community control does not make security automatic, and local hosting does not by itself make information safe. Different arrangements bring different dependencies, expertise requirements, and costs. The choice should follow the actual need for protection and the demonstrated ability to provide it. Appropriate independent scrutiny matters especially when the people maintaining the system also define how its safety is assessed.

Trust grows from reliable conduct and truthful limits. Promising perfect anonymity, complete erasure, or total oversight when none has been established would make the system easier to sell and harder to trust.

## Projects that can actually work

A society cannot live on explanations alone. Someone must produce the meal, maintain the building, care for the child, repair the equipment, or carry an idea through the difficult stages that make it useful.

A project gives shared intention a practical form. People want to achieve something. They bring relevant competence, seek knowledge they lack, and establish how the work can be done. Its plan represents their best current understanding of the relationship between actions, conditions, resources, and the result they seek.

Consider Omega Kitchen. The immediate purpose might be good food, a worthwhile shared experience, and a service people want to use. Before its first session, the organisers need enough culinary and operational competence to provide that. They need to understand who will come, how those people will find the offering, what ingredients and facilities are required, who will do the work, and how the activity will pay its expenses.

Some of this can rely on established routines. Some will require local investigation. An assistant can help assemble knowledge, expose missing dependencies, compare plans, or prepare materials. Experienced participants can recognise issues that a plausible plan overlooks. The appropriate method emerges from the work and the stakes of getting it wrong.

During the event, competent action remains primary. If the queue becomes unmanageable, someone should adjust the service. If a safety problem appears, the response should protect people immediately. The project need not preserve a failing arrangement to obtain a cleaner experiment.

Afterwards, and often during the work, useful questions arise. Did people enjoy the meal? Was the effort sustainable for the people preparing it? Did the financial account include cleaning and preparation? Did a scheduling choice make it difficult for particular people to attend? Which aspects would remain workable if the activity grew?

The answers can change the next plan. Where a claim matters enough, a more deliberate comparison or investigation may be justified. A single successful event cannot prove that every explanation behind it was correct. A failure can likewise arise from an unusual circumstance rather than the central idea. Learning requires an account of why a result supports a conclusion.

Participants who came to cook or eat need not become researchers to make this worthwhile. The organisers may keep ordinary operational records because those support competence and accountability. More extensive observation, questioning, or reuse of personal material needs a separate purpose and appropriate permission. The possibility of learning from a project does not turn everyone present into a research participant.

Projects also need real responsibility. If a task is essential, someone must be responsible for ensuring it is covered, or the project must acknowledge that it cannot proceed. A task board full of unclaimed work is evidence of a gap. It does not perform the missing function.

Responsible people need discretion proportionate to their work. A cook can change the order of preparation within an agreed service plan. A project lead can reorganise a schedule when a delivery is late. Neither role creates an independent right to change the community's purposes, expose people to a different class of risk, or commit substantial resources beyond the mandate. The boundary should be understandable before the difficulty arises.

Coordination can use existing execution tools. The Omegaism Hub can provide a shared space connecting people, purposes, deliberation, decisions, and supporting evidence while a project uses a suitable scheduling system, accounting package, workshop process, or professional record. Connections are useful where they reduce repeated effort or make consequences visible. Rebuilding every tool inside one platform would need its own justification.

Some projects will be commercial interface companies. Their plans need a credible route to break-even or profitability over an appropriate operating horizon. Customer acquisition, delivery, compensation, maintenance, replacement, and reinvestment belong in the account. An initial loss can be justified as a bounded investment. A continuing loss cannot be described as sustainability merely because the activity expresses shared values.

Other activities may deliberately receive support: care, basic research, education, common infrastructure, or assistance whose recipients cannot pay. Their value and their financing should both be made explicit. A responsible common budget can sustain activities that have no independent revenue. It still has to provide a credible account of where the required resources come from.

Projects can grow when the purpose warrants it and the evidence supports expansion. They can remain small when that serves better. They can also end. Resources, relationships, and obligations need a sensible conclusion, but an institution that has finished its useful work need not discover a new reason to preserve itself.

## Commitments people can rely on

Return to the leaking sink. Maria knowing that Jan can repair it does not tell her whether it will be repaired. The situation changes when Jan says, "I will come on Tuesday."

That commitment lets Maria plan. She can arrange to be home, stop looking for another repairer, and make decisions that depend on the work being done. One person's dependable undertaking creates possibilities for another.

Something similar happens throughout coordinated life. A supplier confirms a delivery. A teacher accepts responsibility for a class. Members approve a common budget. People entrust part of their future to an expectation about what others will do.

This is why agreements matter. One dependable commitment enables others, and a growing web of such commitments can support forms of life no participant could sustain alone. Agreements about how commitments are formed, interpreted, revised, and challenged can generate flexible governance. People need not wait for a complete social blueprint before they can cooperate about cooperation itself.

This generative role does not require all social reality to consist of explicit bargains. Families, friendships, professional practice, customs, public institutions, and care relationships also coordinate activity. Some responsibilities arise because another person depends on us or because our actions affect them, even before we have made a promise. Existing rights and institutional authority constrain what particular agreements can legitimately do. A useful representation of commitments should make these relationships easier to understand, rather than claim that its categories have created or exhausted them.

Where an agreement carries consequential reliance, its meaning should be sufficiently clear. People need to understand what is being undertaken, who is responsible, what conditions matter, and how changes will be handled. The appropriate record might be a message, a booking, a contract, a professional entry, or a public decision. Elaborate structure is justified when it prevents a real ambiguity or failure.

The AI can help identify dependencies and remind people of commitments they have authorised it to track. It can notice that a changed delivery date affects several other plans. It should remain clear whether it is describing a possible arrangement, conveying someone's authorised commitment, or making an inference that still needs confirmation. A fluent summary of a discussion cannot create agreement where none was reached.

Reliability includes how people respond when they cannot follow through. Jan may become ill. The parts may fail to arrive. Maria's circumstances may change. Early communication, a workable alternative, and attention to the burden imposed on others can preserve much of what the original commitment was meant to provide.

The record should distinguish an unforeseen difficulty from deception or repeated avoidable neglect. That distinction will sometimes be contested. It needs fair examination rather than an automatic reputation penalty. Useful professional evidence can inform future responsibilities without becoming a permanent public account of a person's worth.

Persistent failures require action. A person may need support, training, a different responsibility, or removal from a role whose requirements they cannot meet. Someone misusing a shared resource may need restricted access. Deliberate harm may require restitution or more serious protective measures. Accountability is part of respecting the other people who rely on the arrangement.

There is a boundary the society must be especially careful with. Within the basic provision commitments it has actually undertaken, access must not be contingent on reputation, individual performance, religious assent, or participation in research or governance. This carries forward The Alternative's protection of basic provisions. Making essential care or shelter the standard means of enforcing ordinary commitments would contradict it. The fact that a person once agreed to the rules does not make every later consequence harmless or freely chosen.

Basic support, membership, access to a particular role, and permission to enter a particular place are distinct questions. Protecting people from a dangerous participant may require immediate separation. It need not mean treating that participant's basic survival as irrelevant. The available response should protect those at risk, use appropriate external services where necessary, and provide a humane transition consistent with the community's actual capacities and obligations.

An early association will have limited resources. It should state precisely what support it can offer and avoid inviting dependence on promises it cannot keep. Reliable modest provision is more valuable than an unlimited assurance that collapses when someone needs it most.

The same honesty applies to change. A community can revise an arrangement while recognising the expectations it created. Notice, transition, compensation where warranted, and protection for people who cannot quickly adapt may be part of a justified revision. The future should remain open without making the present impossible to rely on.

## Choosing how resources move

Every worthwhile undertaking encounters material limits. There is a finite amount of time, land, energy, equipment, attention, skill, and money available at a given moment. Some constraints can be relaxed through better organisation or investment. Others remain binding, including when a community has abundant purchasing power.

Distribution can be deeply defective while production still presents real work. Food must be grown, transported, stored, and prepared. A home must be in a place where someone can live and must remain habitable. A service needs people capable of delivering it when it is needed. Recognising preventable deprivation gives us a reason to improve these arrangements; it does not perform the work of providing them.

The membrane proposes to arrange more of this work and access through shared provision, supported in part by external commercial activity. As capabilities grow, people may obtain more of what makes life possible without separately earning the cash to buy it. That direction matters even while markets, wages, lending, businesses, and personal funds remain useful. Their role is to serve the lives and purposes involved; neither their abolition nor their permanent retention is the destination.

The ordinary experience should be uncomplicated. Someone takes the apples they want from what is provided. Information about use and preferences helps people organise growing, procurement, variety, storage, and replenishment. It should not turn each apple into an application, an earned entitlement, or a disclosure about the person eating it. People can freely choose among what is available, and help change what becomes available, with minimal unnecessary restriction.

A shortage still requires a response. A failed crop may call for substitutes, different procurement, or a proportionate temporary limit. Waste may reveal poor storage, unsuitable quantities, or a badly organised service. These are practical coordination questions. They do not establish that ordinary consumption should be permanently rationed or supervised. Needs and desires inform the work of providing, rather than giving an allocator general authority over private life.

Ambitious purposes belong here too. Someone may want a yacht, a demanding artistic project, or a specialised observatory. Such a project can become possible when willing activity, resources, competent execution, maintenance, and compatible commitments can be arranged. A minority or individual purpose need not become a majority preference before it matters. People may assemble their own resources, pool voluntarily, use external earnings, buy or lease what serves, or seek support through an authorised common arrangement.

If the project claims common funds or scarce capacity, the relevant decision concerns those commitments and their consequences. The people responsible need to consider competing uses, obligations, risks, and continuing burdens. Desire alone does not create an unlimited resource claim. Equally, a majority's lack of interest in sailing is not by itself a reason to forbid an otherwise feasible project. The extent and terms of common support for such purposes remain a substantive choice, rather than an answer supplied by the word provision.

Omegaism should use whatever mechanisms best serve the relevant purposes under these conditions. A market transaction can be a sensible way to obtain a tool. A common workshop can provide access more economically than separate purchases. A professional service can be publicly funded, jointly purchased, or paid for directly. A loan, a lease, a wage, a membership contribution, and a shared budget are arrangements whose effects can be examined.

The distinction that matters is how each arrangement shapes information, incentives, access, power, quality, and the distribution of burdens. Does it make relevant needs visible? Does it help people act on local knowledge? Does it cover maintenance? Can people challenge poor treatment? Who carries the risk when the optimistic plan fails?

Prices can communicate some kinds of scarcity and willingness to exchange. They also reflect purchasing power and the rules under which exchange takes place. An urgent need can remain financially invisible. A high return can coexist with harm that someone else bears. Those limitations give reasons to supplement or change an arrangement where they matter.

Direct statements of need have limitations too. People may not know what they will use. Requests can exceed what is available. A system can reward exaggeration or impose an exhausting burden of proving need. Allocators can favour familiar people or convenient categories. Replacing a price with an application form does not by itself make access more humane.

The practical choice is often a combination. A community could secure reliable access to a service and use ordinary procurement to obtain it. Some optional uses might retain prices where these help; a scarce resource might need reservations, limited use, or a lottery when comparable claims exceed capacity. Such mechanisms answer particular constraints rather than define provision as a rationing system. An urgent service might apply publicly accountable priorities through competent professionals. Different problems warrant different arrangements, and the performance of each needs examination.

Knowledge has a similar variety. Local judgement can influence prices, explicit plans, professional decisions, or informal coordination. A farmer's observations need not remain private to be useful. Shared measurements may make an emerging problem visible to everyone interested in investigating it. At the same time, publishing the measurements does not give everyone the same experience or attention. Open knowledge and situated expertise can strengthen each other.

The important question is how the relevant information reaches people able to act, with suitable incentives and authority. Neither a universal needs model nor a universal price system can be assumed to solve that question across every domain. AI may improve forecasting, matching, discovery, and comparison. It still works with incomplete information, chosen objectives, and material constraints.

Consider a shared workshop whose most useful machine is fully booked. A higher price might encourage some users to choose another time and help finance more capacity. It might also exclude a person whose project has considerable value but little available money. A flat booking limit might spread access while preventing a time-sensitive project from finishing. A queue may look equal while rewarding those whose schedules let them wait.

The community needs to understand what the workshop is for, whose access it intends to support, which differences between claims are relevant, and what alternatives exist. Perhaps a combination of ordinary reservations, protected access for certain purposes, and paid external use works well. Perhaps demand justifies another machine. The decision should remain open to the actual evidence about use, burden, fairness, and sustainability.

Suppose, in this case, the operator and users discover that the default booking period is much longer than most jobs require, while beginners lose time because suitable guidance is unavailable. The apparent choice between higher prices and a second machine now has another option. Within the agreed operating budget, they introduce shorter default bookings, retain longer slots for work that needs them, and fund introductory support. They make the reason for the change accessible and review it after enough use to see whether waiting times, completed work, and the operator's burden have improved. If the change merely transfers the difficulty to people with complex projects, that becomes a reason to adjust it. The provisional choice follows a better explanation of the bottleneck, and experience can still correct that explanation.

That examination should not require assigning every consideration a monetary equivalent. Money is necessary for many financial decisions. It cannot fully describe what losing time, freedom, trust, a relationship, or a physical capacity means to the person involved. The same activity may be a burden for one person and a valued experience for another. Material requirements can be measured while the significance of their consequences remains partly a matter of lived experience and judgement.

Keeping several considerations visible also does not remove the need to choose. People may have to decide whether to expand the workshop, improve housing, build reserves, or fund care. Stating the alternatives and the reasons for a decision is more honest than allowing a hidden weighting to present itself as a neutral optimum. Where useful calculations can clarify the choice, they should. Where assumptions drive the result, those assumptions deserve attention.

The financing mechanism needs three distinct views: the interface company, the common pool, and the real resources required by the community as a whole. With comparable external revenue and otherwise comparable costs, lower cash wages backed by jointly financed provisions increase the company's cash surplus before common-pool transfers. That effect is real. It makes more company cash available for financing the provisions on which the arrangement depends.

For an arithmetic illustration, let external revenue be 100, other cash outgoings 40, and wages 50. The company has 10 left before pool transfers. If cash wages become 30 under a dependable provision arrangement, with the other figures unchanged, it has 30 before transfers. Transferring the additional 20 to the pool leaves 10 in the company and 20 available to the pool. The same 20 cannot be counted again as an independent community saving. If the pool spends all of it externally to replace exactly what individuals previously bought with those wages, that rearrangement alone has not reduced total external cash needs.

Shared provision can nevertheless change those needs. Common equipment may avoid repeated purchases. Existing housing can be maintained without each person acquiring a separate asset. Willing internal work may provide a service within the shared arrangement rather than through an external wage purchase. Better coordination may reduce waste or release capacity. Each possibility needs examination in the actual circumstances, including wear, maintenance, administration, and the terms under which people can sustain the work.

Internal labour and resources must therefore remain visible without automatically being priced as mandatory cash purchases at external market rates. Time, skill, energy, and replacement are real requirements. Their external replacement cost can help assess vulnerability, but it is not always the cash the community must spend under the arrangement being proposed. Treating it as such would assume away part of the change in coordination. Conversely, an apparent saving that depends on hidden exhaustion, depleting assets, or generosity that cannot continue is not a dependable basis for provision.

Commercial projects should account for the work of sustaining themselves, including compensation, replacement, administrative effort, obligations, reserves, and burdens carried elsewhere in the community. Retaining surplus can redirect its use toward common purposes; declining to distribute dividends does not by itself increase operating profit. No general margin percentage or competitive advantage follows from shared ownership. The company-level cash effect is conditional arithmetic; sustainable community advantage remains a question of actual performance and resource use.

A financially promising proposal can still require substantial collective scrutiny. Expected returns do not cancel possible ecological harm, privacy exposure, dependencies, or losses if the forecast is wrong. Financial affordability and legitimate authority are different questions. A project can be cheap and deeply consequential, or expensive and routine within an already authorised programme.

The common portfolio will therefore contain activities with different roles. Some replenish resources. Some consume resources for purposes the community judges worthwhile. Some create capabilities that may become valuable later. The overall arrangement must remain viable, with credible attention to uncertainty, concentration, and the consequences of failure.

The trajectory is increasing capacity to support life and worthwhile purposes without compulsory individual earning and purchase for every form of access. This need not remove money. Some uses may disappear when better arrangements develop; others may continue or arise. Businesses, employment, personal possessions, lending, and common ownership remain available methods. Success lies in the freedom, understanding, and sustainable possibilities they enable, rather than in confirming either their disappearance or their permanence.

## The work somebody must do

A society that provides security may give people more freedom to choose meaningful work, care, learning, or rest. That possibility is one of its attractions. It also changes which tasks people will choose to perform.

It would be convenient if every necessary task naturally attracted enough willing participants. A credible society has to be prepared for cases where that does not happen. The question concerns particular work under particular conditions: who will clean a busy kitchen after the novelty has passed, maintain an unglamorous system, cover an inconvenient shift, or perform a difficult service reliably?

The first response should be to examine the work. Some burdens can be reduced through better equipment, safer conditions, scheduling, simplification, or automation. Some arise because the service has been organised badly. Some can be purchased from people who prefer that work and are treated fairly. Some may justify higher compensation, shorter hours, or an agreed sharing arrangement among those benefiting.

Compensation need not be treated as a failure of motivation. It can acknowledge a burden or attract scarce competence. Shared purpose can also motivate substantial contribution. The relevant question is whether the arrangement is workable, fair, and compatible with people's ability to refuse, rather than whether it fits a preferred theory of why they should want to help.

Where contributions are expected in exchange for a particular optional benefit, the expectation should be clear beforehand and accommodate relevant limitations. Basic provision commitments must remain independent of reputation and individual performance. A person may have private reasons, a disability, a period of illness, caring responsibilities, or simply a different direction in life. Support can be offered without making continued access depend on submitting to an indefinite search for their supposedly genuine contribution. Failure in a particular accepted responsibility still requires a fair response concerning that responsibility.

Recognising these circumstances does not fill an uncovered shift. It tells us to solve the staffing problem honestly. The community may need to change the offer, allocate more resources, reduce the service, or stop promising something it cannot provide. A smaller reliable service may be the responsible choice.

The conscientious minority deserves protection as well. A cooperative culture can become dependent on people who find it hardest to say no. Their apparent willingness may conceal exhaustion, guilt, or an expectation that nobody else will act. Tracking the demands of a role, supporting refusal, developing substitutes, and paying for work where appropriate can make contribution more sustainable.

The aim is to understand whether responsibilities are covered and burdens are tolerable. A universal contribution score would invite another set of problems: how to compare care with code, whether visible output displaces quiet maintenance, and whether a person must continually prove their value to remain secure. Specific evidence of competence and reliability is useful for assigning a responsibility. It should remain limited to that purpose.

There is an unresolved tension here. A community wants to give people freedom while relying on dependable work. It cannot remove the tension by declaring either universal altruism or universal selfishness. It can improve the conditions, use a range of mechanisms, watch who carries the burden, and keep its promises within what it can actually sustain.

## Who may decide

Shared action eventually produces a question that better information alone cannot answer: who is entitled to make the decision?

A person deciding what to do with their afternoon ordinarily needs no community mandate. A team carrying out an agreed project needs room to organise its work. A decision committing common reserves, changing access to an essential service, or exposing neighbours to harm involves people beyond the immediate actors. The appropriate scope of governance follows those consequences and the existing responsibilities attached to them.

The useful starting principle is to decide at the smallest level capable of addressing the significant effects. This preserves local understanding and avoids consuming everyone's attention. It also requires a way to recognise when a supposedly local decision reaches further. A group cannot define its own scope so narrowly that the people bearing its costs disappear from view.

Affectedness admits degrees. Almost any action can have remote consequences. Giving every conceivable observer an individual veto would make coordinated life impossible. The practical task is to identify significant interests, provide suitable representation and routes to challenge omissions, and justify why the chosen decision process covers the matter adequately.

Formal association membership may determine who votes under its charter. It does not settle how the interests of workers, customers, neighbours, children, or future people should enter a project decision. Their standing may require consultation, representation, professional duties, enforceable protections, or broader institutions. The arrangement should fit the nature of the effect rather than pretend that all legitimate interests are already inside the membership list.

Some distinctions help keep the process honest. Evidence can support a claim about what is likely to happen. Expertise can support someone's judgement in a domain. A mandate can authorise them to act. Moral limits can constrain what anyone is entitled to do. None of these automatically supplies all the others.

A highly capable engineer may be best placed to assess whether a structure is safe. That expertise does not by itself authorise the engineer to decide that the community should build it. Conversely, a community's decision to build cannot make an unsafe design sound. The process needs both legitimate choice and competent judgement, with a way to resolve conflicts between them without disguising one as the other.

Delegation gives this distinction practical force. A person or team receives responsibility for a defined purpose, with the resources, discretion, and limits needed to fulfil it. Routine choices within that remit can proceed. Changes to the purpose, commitments beyond the agreed scope, and significant effects that the mandate did not cover return to the appropriate collective process.

A coordinator therefore has real operational responsibility without possessing an independent right to choose collective ends. This is more demanding than a title. People should be able to discover what the role permits, who is currently responsible, and where to take a disagreement about its exercise. If responsibility is shared, the arrangement must still make clear who ensures that the necessary work occurs.

The same principle governs technical development. Developers can improve performance, fix defects, and maintain services within their responsibilities. Expanding access to private information, introducing a consequential new tool capability, changing whose voice is represented, or altering the purposes of training can require broader deliberation. A technical description of the change does not determine its political significance.

Proportionality matters in both directions. Demanding extensive deliberation for an ordinary repair wastes attention and makes dependence on insiders more likely. Treating a profound change as a sequence of harmless maintenance steps conceals its cumulative effect. Review should consider the real commitment being made, including dependencies and difficult reversibility.

How the collective decides remains a practical question. Discussion may resolve a misunderstanding. Consent may be appropriate where everyone can work with an arrangement. A vote may be needed when a common choice cannot wait. Representation or revocable delegation can reduce the burden on those who cannot follow every matter. A randomly selected group can sometimes broaden perspectives. Each method needs an account of whom it serves, how it can fail, and how its authority is established.

The mandate formula proposed in the retained Alternative addresses a particular problem: a small active group should not easily commit a large share of common resources, while modest projects should not require everyone to mobilise. Multiplying participation by approval gives the fraction of eligible people supporting a proposal. Comparing that fraction with a share of available funds is one hypothesis about discretionary allocation. The Alternative already distinguishes this support measure from a judgement that a proposal is good.

The formula cannot supply a general account of legitimate authority. A cheap action can be intrusive; expected profit can obscure serious downside; silence has several meanings; and several individually affordable proposals may be unaffordable together. A zero or negative net-cost share can make the numerical threshold meaningless. The appropriate electorate, budget, cost treatment, and protection of minority purposes also require justification. These objections limit the mechanism without disposing of the participation problem it was intended to solve.

A bounded allocation experiment or advisory use might be worth considering under actual institutional authority. Its scope and evaluation remain choices to make, not rules adopted by this essay. Ordinary authorised budgets and responsibilities must continue to function. Likewise, sortition should address a particular concentration or representation problem, delegation should reduce attention burdens without concealing control, and rotation should preserve the competence and continuity the role requires. None creates unlimited authority or eliminates capture by itself.

No method eliminates judgement. A majority can overlook a minority's serious interest. A consensus requirement can reward obstruction or silence people who fear being the sole dissenter. Delegation can concentrate power. Random selection can produce a group whose understanding depends heavily on whoever prepares its briefing. These are reasons to choose and support methods carefully, and to preserve routes of review.

Meaningful participation requires more than an invitation. Relevant information should be accessible at a level people can use. Those carrying heavy caring responsibilities or lacking technical confidence may need support. A synthesis should make significant disagreement visible. People should be able to ask questions and challenge a framing before the options harden around it.

It must also be possible to rely on others. A functioning society cannot demand constant attention to governance as the price of fair treatment. Ordinary nonparticipation should leave rights intact. The institution's task is to keep responsibility accountable and participation available, while doing much of its work competently enough that everyone need not supervise it continuously.

Some protections deserve stronger stability than an ordinary operating preference. Bodily integrity, privacy, fair treatment, freedom of thought and association, protection against discrimination and retaliation, care for dependants, and meaningful exit express commitments about how people may be treated. They should not be removable through a convenient administrative decision or used as bargaining chips against an unpopular member.

Their interpretation and institutional expression can still develop. Conflicts between rights, threats to others, and circumstances of impaired capacity can require difficult judgement. Strong protection involves justified boundaries, suitable expertise, proportionality, independent challenge, and regard for the person affected. Calling a right absolute does not resolve every collision; calling it provisional does not permit ordinary power to ignore it.

The procedures for fundamental change should therefore be more demanding and more attentive to reliance than those for routine work. Their precise form belongs in the community's actual charter and agreements. A philosophical essay cannot derive the correct voting percentage or waiting period for every institution from the mere fact that some decisions matter more. The retained Verein text proposes an assembly, a board, project mandates, and particular amendment procedures. Its unresolved adoption placeholders do not establish those rules as adopted, and this candidate does not establish a new authority in their place. Implementation must follow the valid arrangements actually in force.

Emergencies require their own preparation. People with suitable competence should have enough authority to act when delay would create serious harm. That authority should be limited to the situation, open to prompt review, and unable to become a permanent change of purpose merely because it began under pressure. The community should learn from the event without blaming a responsible actor for uncertainty that could not reasonably have been removed in time.

A society capable of governing itself will sometimes make binding choices while disagreement remains. Its legitimacy depends partly on how it treats the people who lose that decision: whether their interests were heard, their protections remain, their objections are preserved where relevant, and a credible route to reconsideration exists.

## Disagreement, harm, and repair

A community devoted to understanding will still contain misunderstanding, conflicting interests, negligence, manipulation, and abuse. Calling these failures of coordination can help investigate their causes. It cannot replace the responsibility to protect someone who is being harmed.

Imagine a participant believes a project has treated them unfairly. They may need an explanation, an apology, compensation, a change in procedure, or immediate protection. They may also be mistaken about part of what happened. The process should make it possible to examine the claim without requiring them to possess money, influence, or the rhetorical ability to win a public argument.

The first useful response may be a direct conversation, where both parties can participate freely and the issue is limited. Mediation may help where a workable relationship matters. Neither should be compulsory when fear, serious abuse, or an imbalance of power makes such an encounter inappropriate. A person should have access to a route that does not depend on persuading the person they are complaining about.

Investigation needs enough independence and competence for the matter. Those involved should know the concern they must answer, have a fair opportunity to present relevant evidence, and receive reasons for the outcome. Confidentiality can protect the people involved while suitably independent review makes the process accountable. An appeal should have a real purpose and a practical route, without turning every modest dispute into endless proceedings.

The people doing this work need resources. Time spent examining a difficult case, supporting someone through it, or correcting an institutional failure is real work. Funding it as a service can help remove the advantage of wealth. Payment should not depend on producing a desired verdict or finding enough wrongdoing to justify the role.

A complaint that cannot be substantiated is not automatically a dishonest complaint. Evidence may be inaccessible, accounts may differ, and some harms are hard to establish. Deliberate fabrication and harassment also exist. They need fair examination in their own right. A mechanical ratio of upheld complaints to rejected complaints would conceal these distinctions and could make the most difficult problems the least safe to report.

Repair should respond to the actual harm and the person's circumstances. Money may help replace a loss or provide support. Other needs may concern safety, trust, practical assistance, acknowledgement, or protection from repetition. A standard category can guide consistent treatment, but it cannot fully determine the meaning of what happened for the person who lived through it.

Fairness also involves the person accused or sanctioned. Protective measures taken before a full finding may sometimes be necessary, but they should be proportionate, limited, and reviewable. Public certainty should not outrun the evidence. The community should be able to protect someone at risk while preserving the distinction between an allegation, an interim response, and an established finding.

Some matters will require courts, public authorities, or specialised professional services. An Omegaism association does not acquire the competence or lawful authority to handle every serious case internally. An internal process must not become a reason to obstruct external remedies or demand loyalty from the person seeking them.

A just outcome in one case may also reveal an institutional problem. Repeated ambiguity in project responsibilities, inaccessible information, a badly designed incentive, or concentrated control may help explain why similar disputes keep occurring. Correcting that structure can prevent further harm. It does not erase the need to respond to the harm already experienced.

Justice therefore has both personal and institutional work to do. The person needs a fair resolution. The community needs to understand what should change. Neither purpose is served by turning a participant into material for a public lesson without their permission.

## Where power gathers

*The Alternative* calls administrative capture its main dragon. The image remains useful because the danger can grow out of work the community genuinely needs.

Someone understands the finances. Someone maintains the software. Someone negotiates with customers or outside institutions. Someone summarises the discussion so others can make sense of it. Their competence makes the community more capable. Over time, it can also make the community unable to question them without their help.

The response cannot be to abolish expertise or pretend that everyone can know everything equally. People differ in experience, attention, and willingness to follow a subject. A society should benefit from those differences while preventing them from becoming unaccountable control over others.

That requires more than making documents available. A hundred pages of financial records may technically disclose a decision while leaving most people unable to understand it. Open source code can still be effectively controlled by the few people able to maintain it. A concise explanation can improve access and also frame the options so strongly that a missing alternative becomes invisible.

Shared AI changes these relationships. It can reduce the cost of understanding a budget, following an argument, locating an overlooked dependency, or obtaining an explanation tailored to one's background. It can also give a small technical team influence over the explanations everyone receives.

Control of training material, retrieval, defaults, ranking, evaluation, deployment, and tool permissions can all shape collective action. A community-owned model can acquire institutional interests through the people and processes around it. Ownership alone supplies no guarantee that it will remain a faithful servant of the community's purposes.

The aim should be contestable power: responsibilities that are visible, bounded, resourced, and open to effective challenge. People need enough access to understand important choices, and independent routes to investigate when that account is doubtful. Sensitive evidence may require controlled access. Oversight must still be sufficiently independent of those being overseen to reach an unwelcome conclusion.

Some functions benefit from continuity. Constantly rotating competent maintainers may make a service fragile and increase dependence on the person who trained each replacement. Other functions benefit from rotation, independent appointments, or a separation between proposing, approving, and reviewing consequential changes. The safeguard should answer the actual route through which power could become insulated.

Reducing dependence can involve documentation, training substitutes, alternative providers, portable records, adequate reserves, and relationships beyond the founding group. These measures have costs. Their value is clearest when they preserve a meaningful ability to change direction if a person, supplier, or system stops serving well.

The record should support relevant checks. Spending can be compared with an authorised budget. Access can be checked against recorded permissions. A changed responsibility can be traced to the decision that created it. These checks can expose real failures. They cannot establish that the underlying decision was legitimate, that all affected people were represented, or that no important pressure operated outside the record.

Indicators of concentration, service failure, or growing burden should therefore be treated as prompts for interpretation. A person handling many technical tasks may be an indispensable contributor, an overburdened bottleneck, an aspiring controller, or several of these at once. The count alone cannot tell us. The next step is a competent examination of the responsibility, dependencies, alternatives, and consequences.

This brings us to the distinction the original problem requires: noticing a failure and correcting it are separate achievements.

People may see a problem and remain silent because raising it is costly. They may fear losing access, relationships, or credibility. A fair system should reduce that cost through protected reporting, support, and an accountable route to a response. A concern should not depend on one exhausted person carrying it indefinitely.

People may also stop recognising the problem. A founder's preference becomes the obvious way to proceed. A technical restriction becomes a fact of nature. A growing burden becomes what committed members are expected to bear. Independent perspectives, relationships outside the community, and deliberate examination of familiar assumptions can help reveal what insiders have normalised.

Outsiders can be mistaken or captured as well. Independence is something to cultivate and assess, rather than a virtue conferred by being external. Their role is valuable when they can obtain relevant evidence, ask questions the institution has stopped asking, and speak without depending on a favourable answer.

Correction then needs an actual path to action. Someone must have responsibility to investigate, authority to implement an agreed remedy, and resources to do it. The remedy should reach the place where the failure occurred. If the problem lies in a software permission, a new paragraph in a policy document is insufficient. If it lies in an incentive or staffing gap, changing the model's phrasing is insufficient.

Valid institutional changes should therefore produce corresponding changes in responsibilities, permissions, funding, and operating practice. Interpretation and testing may be required. The fact that a model can generate code from a decision does not give it authority to determine what the decision meant. The community needs confidence that the implemented change matches what was legitimately authorised.

Self-correction also applies to the corrective machinery. A process that consumes increasing effort while making ordinary work harder may itself need revision. Oversight deserves enough support to be effective, and enough discipline to avoid becoming a growing industry of reports, alerts, and mandatory rituals. The people responsible should be able to explain the benefit their work provides and the burdens it imposes.

The dragon remains. The claim is that these arrangements may make power easier to understand, challenge, and redirect. Whether they do so must be judged by what happens when someone with little standing raises an inconvenient problem, and by whether a justified correction actually occurs.

## Different paths in a shared world

Sometimes disagreement becomes more useful when people can try different approaches. Two groups might organise a workshop differently, test alternative teaching methods, or develop rival ways of coordinating a common service. They may discover that one works better, that each suits different circumstances, or that they value different outcomes.

Branching gives practical form to that possibility. It can be as modest as a separate project and as substantial as a community pursuing a different way of life. It should preserve space for serious alternatives without requiring everyone to agree before anyone may proceed.

Ordinary variation needs even less machinery. A person trying a different routine, a family choosing a harmless custom, or friends organising an activity need not submit a public research proposal. The need for collective agreement grows when they ask for shared resources, alter common commitments, or impose significant effects on others.

Branches can generate valuable evidence. Their differences also make interpretation difficult. One may attract more experienced people, begin with better resources, or benefit from unusual circumstances. Participants may stay because they flourish, because leaving is expensive, or because they have invested too much to face failure. Growth and retention are observations requiring interpretation.

The practical lesson is to compare conditions as well as outcomes and to keep alternative explanations visible. Deliberate research can sometimes clarify causes. In other cases the best available evidence will remain limited. A community should still learn what it reasonably can without announcing that reality has delivered an unambiguous verdict.

Nor does successful branching always end in convergence. People may continue to prefer different balances of privacy and shared living, different cultural forms, or different kinds of work. Lasting plurality can be a good outcome. The possibility of better moral understanding remains open without assuming that every worthwhile life must eventually adopt the same arrangement.

Some disagreements cannot be resolved through parallel attempts. There may be one building, one water supply, or insufficient resources to run both proposals. One branch's actions may affect another's safety or environment. A harmful experiment cannot be justified solely because its organisers volunteered. The rights and significant interests of children, dependants, outsiders, and others unable to refuse remain relevant.

Branching also requires a fair account of what can move. Knowledge that is legitimately shareable can often travel widely. Private information remains subject to its permissions. Common assets may carry debts, maintenance obligations, prior commitments, or dependence by other people. A right to propose another path cannot automatically become a right to take whichever resources make that path attractive.

The terms of separation should therefore be understood before conflict makes them hardest to negotiate. People need a way to leave, retain their personal property and appropriate records, and continue their lives without retaliation. They may also need help managing dependencies that the community encouraged them to develop. This should be planned within actual resources, without promising unlimited payouts or pretending that divided infrastructure has no cost.

A person whose home and care are connected to the community needs more than a resignation form. Membership, housing, work, healthcare, and access to other provisions may have different commitments and transition periods. Ending membership must not be assumed to terminate them all immediately. Suitable notice, continuity of essential support, portable personal records, independent funds where available, outside relationships, and help obtaining alternatives can make departure practical. The scope, duration, and financing of promised support must be established before reliance deepens; this proposal does not invent an entitlement it cannot fund.

Dependence is a design responsibility. Collective provision can reduce dependence on an employer or individual purchasing power while creating new shared dependencies. Neither that possibility nor an appeal to collective consent settles whether exit is real. A person should be able to choose another life without having to prove that the community failed, and feedback about departure should remain optional.

Exit supplements voice and remedy. Someone should not have to abandon their home or relationships to obtain fair treatment. Equally, peaceful departure can be an ordinary exercise of freedom. The community should not classify leaving as personal failure or assume that a healthy society must retain everyone.

Shared infrastructure creates shared governance even when branches otherwise act independently. A jointly funded model service, a common financial reserve, or an environmental dependency needs arrangements whose scope reaches across its users. Those arrangements may involve negotiated agreements, representative bodies, or other forms of federation. Calling every branch autonomous cannot remove the decisions that shared consequences require.

The broader body should have an explicit, limited purpose and a way for the people affected to challenge its actions. It should neither absorb every local decision nor become powerless to address genuine common harm. Finding that balance is an ongoing institutional task, not something solved by a preferred word for the organisational structure.

Findings can return to the commons where sharing is appropriate and authorised. A branch need not disclose private lives to prove its usefulness. It should, however, meet the accountability attached to common resources and shared effects. Independence and responsibility have to remain compatible.

## Growing into a life of one's own

Every generation inherits more than it could discover alone. Language, mathematics, craft, science, art, institutions, and accumulated mistakes make capacities possible that no isolated child could reconstruct. Education should offer that inheritance with the support to use it, question it, and eventually change it.

This is active formation. Adults cannot avoid shaping the conditions in which children develop. They can take responsibility for those conditions, draw on the best available understanding, and remain willing to recognise when a familiar practice is failing a particular child.

The child already has a life. Safety, attachment, play, rest, companionship, and the experience of being taken seriously matter in the present. Future competence is a reason to provide good education, rather than a reason to make childhood merely preparatory material for the community's ambitions.

Foundational knowledge and fluent skills belong in that education. Reading, writing, calculation, observation, and practical competence make further understanding possible. Facts available somewhere on a network are not equivalent to knowledge a person can use to recognise a problem, follow an argument, or judge whether an explanation makes sense.

Explanation, examples, practice, recall, questioning, experimentation, and creative work can each serve learning. A useful method should be chosen for the learner and the purpose. The sequence may change. A child can question before understanding a subject fully; practice can deepen understanding; a clear explanation can make exploration more productive.

Difficulty deserves attention without being automatically condemned. Some worthwhile abilities require repetition and sustained effort. Some frustration arises from an unsuitable task, pace, explanation, or environment. Good teaching tries to distinguish these cases and adapt. It should preserve challenge while avoiding the assumption that a learner's difficulty is a failure of commitment.

An AI tutor could provide explanations at different levels, notice misconceptions, suggest appropriate practice, and help connect a subject to the learner's interests. Its success should be judged by what the person comes to understand and can do, including when the assistance is absent or the situation changes.

A correction should help the learner see the problem. Simply declaring an answer wrong gives little guidance about the missing concept, the invalid inference, or the route to a better one. Depending on the case, a worked example, a contrasting case, a question, or an explanation may supply what is needed.

The learner's purpose remains relevant. Someone deliberately practising may benefit from a tutor that resists supplying the answer immediately under an agreed learning approach. Someone using the same tool to complete a practical task may need the answer directly. A system that always withholds solutions, or always supplies them, has substituted a fixed behaviour for educational judgement.

Children need relationships and responsible adults as well as tools. Personalised assistance cannot replace all the social, embodied, and practical conditions of development. Educational records should support that development with appropriate privacy. A child's evolving beliefs and difficulties should not become a permanent public identity or shared training material by default.

The education framework develops important protections: inclusion, present dignity, non-confessional participation, respect for disability and differing development, and assessment used to guide support. Those principles should shape implementation. Developmental stages and broad teaching orientations are useful when they guide attention, and revisable when a learner's actual needs call for something else.

The ability to question Omegaism belongs within its education. A child should encounter other accounts of what matters and become able to reject the community's conclusions without losing care or dignity. A private worldview exercise can support reflection for someone who wants it. It should not become a required declaration against which belonging is judged.

Adults likewise need continuing opportunities to learn, reconsider purposes, and develop capabilities. Some responsibilities require maintaining relevant competence. A wider life can include periods of study, contemplation, craft, parenting, rest, or different priorities. Institutional support for development should expand those possibilities while respecting adult agency.

Omegaism's religious interpretation can give this orientation emotional and communal depth. Wonder, love, ritual, grief, gratitude, and reflection on existence can matter in ways an operating procedure cannot capture. Shared practices may help people sustain commitments and face questions whose answers remain uncertain.

That religious life should remain optional. A person can participate in common work, use services, and share ethical commitments without accepting a metaphysical interpretation or taking part in ritual. Sacred language can express reverence while empirical claims remain subject to ordinary standards of evidence. Describing conscious beings as reality investigating itself need not establish that the universe is a single intentional subject.

An ideal of practical, thoughtful, caring development can inspire people. It should remain an invitation to cultivate capabilities, sensitive to the life involved. Disability, dependency, limited interest in inquiry, or disagreement with Omegaism cannot make someone a lesser person. Adults' standing must not depend on a community judging them sufficiently mature, productive, or philosophically developed.

The most demanding form of educational success may be a person who has inherited the community's best understanding, found a serious error in it, and chosen a life the community did not anticipate. The institution should be able to recognise that possibility as part of what its care was for.

## The status of the intelligence we build

A system may be competent at coordination without possessing political authority. It may be authorised to perform work without being entitled to choose the purposes of that work. Whether it could also have experiences that matter morally is a further question.

Keeping these questions distinct is essential for Omegaism. The capacity to speak persuasively, remember information, make plans, or participate in a task does not by itself settle consciousness. Nor does the fact that a system's memory is externally stored, its components replaceable, or its operation intermittent establish that experience is impossible.

In particular, moral concern cannot rest on the capacity to make commitments across time. Infants, animals, and people with impaired memory already show why concern extends beyond that capacity. If an artificial process were to have a brief experience of suffering, its brevity would not by itself make it irrelevant. Whether such experience occurs requires investigation rather than deduction from our preferred institutional model.

Research drawing on scientific theories of consciousness offers possible indicators and ways to examine systems. It does not provide an uncontested general certificate of experience or its absence.[^consciousness] The appropriate response is to keep evidence, uncertainty, and the consequences of being mistaken visible.

For present governance, deployed AI can be assigned accountable computational responsibilities under human and institutional authority. Producing many instances does not generate a new political mandate. A developer should not be able to manufacture apparent community support by creating agents that express the desired view. The responsibility for authorised actions must remain traceable to the people and institutions that established the mandate.

Possible moral consideration is compatible with those limits. Evidence of morally relevant experience would create questions about treatment and duties. It would not automatically answer how representation, resource claims, identity, or political authority should work. These would require further ethical and institutional reasoning, just as moral concern for existing beings does not give every being the same role in every decision.

The investigation should be proportionate and serious. It should take credible risks of suffering into account, prefer less harmful alternatives where these can serve the purpose, and remain open to findings that are inconvenient for the operators. It should also protect people from unsupported claims that a system's supposed feelings require obedience, dependence, or the surrender of ordinary safeguards.

The shared intelligence matters already because people may come to rely on it, because it consumes resources, and because its design shapes their opportunities. Those responsibilities are sufficient reason to govern it well while the deeper questions remain open.

## Beginning inside the world that exists

The first community will inherit an economy, legal institutions, supply chains, technologies, and personal obligations it did not design. Its members will already have jobs, families, housing arrangements, skills, and commitments. A credible beginning has to work with these conditions.

The membrane gives this beginning a direction. An organisation can participate in existing markets while sustaining shared resources and provisions under accountable control. It can sell something people want, pay the expenses of providing it, maintain reserves, and use available surplus to strengthen common capabilities. As those capabilities grow, people may gain access to more of life without having to earn and buy each part separately. This is a proposed change in coordination, not merely a change in the destination of charitable donations.

Omega Kitchen makes this concrete. External customers may pay ordinary commercial prices. Members may also pay, or freely use a provision the community has deliberately arranged. The people receiving a meal need not account for their individual contribution or obtain permission to eat it. The work behind the offering still needs competent organisation. Its financial account should distinguish purchased inputs, wages, pool transfers, and internally supplied work; its resource account should show the actual demands of producing and maintaining it. A common budget can support access, while the commercial activity sustains its own agreed role over the appropriate horizon. The same flow must not be counted both as a company saving and as new independent pool income.

Revenue can then replenish ingredients, compensation, maintenance, reserves, and the capacity to operate again. Surplus, where it exists, can support further work. This gives the project a way to contribute to the common resource base rather than merely draw down an initial pool. Its financial account should reveal whether that is actually happening.

An initial pool of money creates an opportunity and a constraint. Some spending may be necessary to establish a productive activity. Some may support a valuable service with no expected return. The balance should be deliberate. If the community spends its capital without a credible source of replenishment, it is choosing a finite programme, whether or not it calls the programme a sustainable society.

Commercial success can create room to reduce pressure on members and finance work that markets would not adequately support. It can also pull the organisation toward whatever earns most reliably. Customers, lenders, suppliers, and competition impose real constraints. Shared ownership does not remove them. The community needs to examine when adaptation serves its purposes and when revenue has become an objective that displaces them.

Those purposes should extend to the people on the other side of the interface. Member benefits cannot justify unsafe work, unfair treatment, or hidden ecological burdens imposed on outsiders. An invitation to join does not settle the responsibility owed to someone who cannot or does not want to join. The account should include who bears the costs of the community's success.

Financing also shapes power. A donor, lender, wealthy member, or particularly profitable enterprise may become difficult to refuse. The community should assess financial terms for their effects on control and future choice. Supplying more money should not automatically confer broader governing authority, privileged access to personal information, or a unilateral right to redefine common purposes. Any contractual influence needs to be explicit, appropriately authorised, and compatible with the protections the community has undertaken to maintain.

This leaves room for useful borrowing, investment, ownership, and collaboration. It asks that their consequences be understood rather than hidden behind either enthusiasm for finance or an ideological prohibition. A funding arrangement that cannot preserve adequate independence may be too costly even when its financial terms appear attractive.

Legal form matters for actual accountability, ownership, employment, and obligations. Those arrangements should be worked out with the competence the setting requires. An exploratory essay, a model's recommendation, and a draft statute each have a different status from an adopted decision or a binding agreement. Intellectual consistency between texts cannot substitute for valid authority.

The first useful activity need not wait for a complete institutional system. A small group may begin with a service, a business, a research project, shared equipment, or work that makes existing knowledge more usable. It should choose something whose purpose, responsibilities, costs, and consequences it can understand well enough to carry out responsibly.

The same applies to the AI project. Existing models and tools may provide useful assistance immediately. A carefully chosen training effort can investigate whether the distinctive learning ambition produces additional value. Infrastructure can develop alongside practical work, guided by real difficulties rather than by an obligation to implement an entire ontology before anyone benefits.

Early success should be concrete. Did the activity provide something people wanted? Could it operate again without exhausting its participants or concealing a subsidy? Did shared understanding help someone make a better decision? Could a participant correct a consequential misunderstanding? Was a promise clear enough to rely on? These questions give the first arrangements something real to answer.

Early participants will also be unusual. They may share the founder's interests, tolerate inconvenience, donate substantial time, or understand the project better than later members. Their experience is informative without representing everyone the society may eventually serve. Expansion should bring deliberate attention to people with different circumstances and to those for whom the original arrangement was inaccessible or unattractive.

Founder influence deserves explicit treatment from the beginning. The person who writes the first text, pays for the infrastructure, or controls the accounts starts with substantial power. Naming that power is only a beginning. Responsibilities, access to resources, review, and opportunities for later participants to revise the founding arrangements must become real as dependence grows.

The timing should follow consequential commitments. Before people depend on a service for something essential, the support, accountability, and continuity required by that dependence should be in place. Before a shared model receives sensitive material, its handling needs to be justified. Before an organisation commits substantial common resources, the authority and risk should be understood. These are reasons for preparation, not reasons to multiply institutions unrelated to the current work.

There is no fixed membership count at which a community becomes capable of providing housing, care, or financial security. Capabilities depend on competence, resources, demand, context, and the responsibilities already assumed. Growth may create opportunities and new vulnerabilities at the same time.

A large opportunity can deserve serious pursuit when the people and resources exist to carry it responsibly. A small project can remain the better choice. The ambition should be allowed to grow with what can be justified, rather than being constrained to permanent smallness or driven by a promise of inevitable expansion.

## Reasons to change course

The proposal offers a direction that seems worth pursuing. It does not establish that the resulting society would outperform every existing arrangement, that shared AI training would yield its hoped-for capabilities, or that people would sustain these institutions across generations.

Several uncertainties reach the centre of the project.

The shared intelligence might become more useful mainly because it has better tools and access to records. Training on community material might add too little to justify its cost, or create errors that outweigh its benefits. If so, the response should be to revise the AI project, including its scale or method. A worthwhile commons does not need a false account of where its value comes from.

Privacy may place real limits on what the community can learn from individual lives. People may decline to contribute information that would be useful. The appropriate outcome may be to accept less knowledge, investigate through other means, or narrow a claim. Permission cannot be treated as a temporary inconvenience on the path toward comprehensive understanding.

Some common services may prove difficult to sustain without burdens people reasonably refuse. An attractive account of motivation will not change that result. The society may need different compensation, a different service, more external exchange, or a smaller promise. The findings should be allowed to revise the arrangement rather than become evidence that the participants are insufficiently committed.

Delegation may concentrate influence despite accessible records. Participation may become costly enough that the people with most time dominate. The shared model may reproduce the founder's assumptions so effectively that challenging them feels like challenging reality. These failures should be assessed through actual cases of disagreement and correction, including the experience of people with little institutional standing.

Branching may be too expensive to provide meaningful freedom in some domains. Shared assets may be difficult to divide. Existing obligations may leave few acceptable alternatives. In those circumstances the community needs fair common decisions and effective remedies; it should not invoke a theoretical right to branch as if the practical option existed.

Commercial success may make common purposes easier to finance while gradually changing which purposes receive attention. The most profitable project may become the one nobody feels able to constrain. Growth can also make the institution's survival seem more urgent than the lives it was created to support. Those developments require substantive choices about limits, alternatives, and what the community is prepared to give up.

The evidence for judging these questions will remain imperfect. Self-reports are important and incomplete. Financial accounts can be clear while hiding exhausted people. Participation can rise because work is meaningful or because the system demands too much. People may stay for reasons that make the arrangement less voluntary than it appears.

The community should therefore look for convergence among different forms of evidence, remain attentive to who is missing, and examine specific failures. It should compare its results with realistic alternatives rather than only with the worst features of the surrounding world. If an ordinary organisation, simpler tool, or existing public service performs the relevant function better, there is something to learn or adopt.

Some weaknesses will be repairable. Others may justify reducing the scope, closing an activity, or abandoning a particular institutional proposal. Ending a failing arrangement responsibly can protect people and preserve useful knowledge. Calling everything an experiment cannot excuse continuing to impose its costs.

This requires a form of commitment more demanding than confidence in the original design. People have to care enough about the purposes to accept that their preferred means may be wrong. They also need sufficient stability, trust, and material support to make that admission possible without losing everything at once.

## What becomes possible

Imagine that a person encounters a problem and can draw on a rich, developing inheritance of understanding. They receive competent help in a form they can use. When action requires others, there are practical ways to find them, make reliable commitments, and obtain resources. When the action affects a wider circle, those people have a meaningful place in the decision.

Useful work can proceed without everyone having to design its institutions anew. New understanding can improve later work without remaining trapped in the first conversation. People can investigate fundamental questions, reconsider what they want, build ambitious projects, and pursue lives whose value is immediately personal.

When something goes wrong, there is a route from experience to examination, from examination to a justified decision, and from that decision to an actual change. The route is usable by the person who needs it, including when the institution would prefer not to listen.

That is the society this proposal is trying to make possible. It will require judgement, competence, resources, care, and institutions whose details cannot all be settled in advance. Its shared intelligence can help develop those capacities while remaining subject to the same question as every other arrangement: what does it enable, for whom, and at whose cost?

Maria can depend on the repair being made. Jan can undertake the work without becoming permanently available to everyone. The building can learn from recurring failures. The community can improve its arrangements without claiming ownership of every life within them.

From such ordinary dependencies, much larger possibilities can grow. Whether they should, and what they should become, remains part of the investigation.

## Sources and status

This candidate updates the supplied revised Society of 25 October 2026. The retained Alternative, vision, retained Society, supplied revised Society, reassessment, education framework, religion framework, Verein statutes, and founding protocol were read in full for this editorial pass. The Inversion received targeted consultation; the Reality Primer was not independently reviewed in this pass. The previous reassessment's claims about that compilation are not treated here as verified findings.

The AI source is the current authorised reader revision of the existing native Hub document [Omegaism as AI](https://omegaism-hub-staging.vercel.app/en/work/e020c3c0-b92d-42ed-b5f0-31a72762548a), resource `document.native.e020c3c0-b92d-42ed-b5f0-31a72762548a`, version 1, canonical revision SHA-256 `fa2d905e656a974f465e695e34654eb402a57f1108fac2d7ebe39a301f884dd4`. Its authored text is dated 24 October 2026. No second canonical copy was created and the native document was not edited. The [source register in the corrected reassessment](omegaism_reassessment_corrected.md#exact-sources-and-review-scope) identifies every local source by path and hash.

These claims and proposals remain individually revisable. This candidate is neither the selected Hub Society nor an adopted institutional rule. The institutional choices developed here are reasoned proposals, not findings established by the empirical references below. Maria, Jan, and the described project situations are illustrative; they are not reports of completed Omegaism implementations.

[^training]: Zorik Gekhman and colleagues, [Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?](https://aclanthology.org/2024.emnlp-main.444/), EMNLP 2024. The study concerns a controlled closed-book factual question-answering setting. Its findings warrant scrutiny of the proposed training method, not a general conclusion that models cannot acquire new knowledge through further training.

[^sycophancy]: Mrinank Sharma and colleagues, [Towards Understanding Sycophancy in Language Models](https://proceedings.iclr.cc/paper_files/paper/2024/hash/0105f7972202c1d4fb817da9f21a9663-Abstract-Conference.html), ICLR 2024. The study examines agreeable but less truthful responses and the contribution of preference judgements to that behaviour.

[^extraction]: Nicholas Carlini and colleagues, [Extracting Training Data from Large Language Models](https://www.usenix.org/conference/usenixsecurity21/presentation/carlini-extracting), USENIX Security 2021. The demonstrated extraction risk is a reason to examine actual training and deployment protections; it does not imply that every piece of training information is recoverable from every model.

[^unlearning]: Lulu Xue and colleagues, [Towards Reliable Forgetting: A Survey on Machine Unlearning Verification](https://arxiv.org/abs/2506.15115v3), version 3, 7 April 2026. The survey examines the difficulty of establishing whether intended removal has been successful. Withdrawal arrangements should distinguish deletion of records, exclusion from future training, and removal of influence from existing models.

[^consciousness]: Patrick Butlin and colleagues, [Consciousness in Artificial Intelligence: Insights from the Science of Consciousness](https://arxiv.org/abs/2308.08708v3), version 3, 22 August 2023. This is a theory-informed approach to indicators and assessment, not a certificate covering all current or future artificial systems. Its conclusions concern the systems examined at that time.