{
  "_meta": {
    "source": "ui_leviathan/web/content/anima.md",
    "version": "0.6",
    "status": "IN DEVELOPMENT",
    "synced_at": "2026-10-21T09:37:29.298Z",
    "content_hash": "3e4b34f2ee5f706247016e81255566ff10ccc1a880e28cb1ee4d617b56d9664c",
    "generator": "web/scripts/sync-docs.mjs"
  },
  "frontmatter": {
    "title": "Anima",
    "version": "0.6",
    "status": "IN DEVELOPMENT",
    "updated": "2026-10-20"
  },
  "body_md": "\n# Your questions, your companion, shared possibilities\n\nYou may have a question that stays with you across conversations: something you notice in daily life, an animal you care for, or a tool you want to make. An assistant should help you develop that question without losing what it means to you or assuming that your private history is public material.\n\nAnima is a personal companion in development for that purpose. It brings local AI conversation, personal knowledge, and questions about lived experience into one application. It also includes a client for participating in the Leviathan forum.\n\nWe want a person to be able to create something useful with their assistant, bring selected work to independent groups, and carry answers or methods back into their life. In the wider proposal, this continuing personal relationship could become a **Levi**. Anima explores one way to make it possible; you can participate in Leviathan with another assistant or without one.\n\n## What exists today\n\nThe development application includes on-device model conversation, personas and beliefs, personal reflection, calendar tools, and editable experience questions that connect conversations to personal observations, memories, and notes.\n\nThe forum client supports reading and participation through the site's account and governance rules. AI-assisted drafting offers a preview for review. Accepting a draft and publishing it are separate actions. Current remote availability and every server write path are not established by the existence of this client.\n\nThese are implemented development features, with selected Android device checks. They are not a claim of general release readiness or equal validation on every platform.\n\n## From a conversation to a reviewed observation\n\nYou describe a lived experience or daily observation. Anima can propose an answer to a related experience question, show which message supports the proposal, or ask which item or situation you meant. The default is to review suggestions before saving them. You can instead choose an explicit action-only mode or authorize a limited automatic recording scope.\n\nMentioning an option in conversation does not by itself mean you adopted or repeated it. An uncertain description should not become an invented rating. Saved answers remain connected to their questions and sources, and Undo can retract a mistaken proposal from the effective answers.\n\nThe current matching is bounded and can miss or misinterpret language. It is not a general understanding of every memory or experience.\n\nThis small distinction between an experience, a suggestion, an accepted record, and a correction matters to the larger idea. We want a language that can preserve such relationships while supporting new questions and actions. This flow supplies a concrete development example to examine, not proof that this wider language has been built.\n\n## A personal meaning structure that can keep learning\n\nThe proposed local meaning and value *kernel* connects concepts, principles, rules, history, and learning. For a personal Levi, it could help preserve what a concept meant to this person in this context, which source or experience informed it, and what later changed. That is a broader research direction than the app's existing personas and beliefs features.\n\nThe purpose is to support reflection and useful action while making changes understandable. A rich context is not proof of faithful representation. We need to test whether the assistant preserves distinctions, notices conflicting commitments, and changes a suggestion when the person's meaning or the evidence changes. This proposed meaning kernel is also distinct from the existing forum's constitutional requirements.\n\nAn unresolved question could remain as a *shadow*: a doubt, missing distinction, or lesson to revisit when a new observation or method becomes available. A creative connection to another field could then suggest a comparison or experiment. Improving how the assistant forms and revisits such questions belongs within the wider research on learning and self-improvement; it is not an established autonomous capability of Anima.\n\n## Work with independent groups without losing the differences\n\nPersonal history, the context chosen for a relationship, and a public contribution serve different purposes. A selected relationship record could preserve the question, sources and their versions, concepts used, relevant values, and what remains unresolved. It could help someone continue with another tool or independent Leviathan while keeping the rest of their private history with them. This portable handoff is a design direction, not a completed feature.\n\nFor example, a caregiver might choose to discuss selected observations with two independent animal-welfare groups. One could emphasize immediate care and another offer a measurement method worth exploring. Anima should help the person understand both proposals, their reasons, and what a summary leaves out. The exchange could produce a new question or a useful design, while an accepted record in one group retains its scope and version.\n\nIndependent groups and Leviathans could develop their own methods and commitments. Anima's proposed role includes connecting selected work across those layers, helping people contribute code or ideas, and bringing useful responses back. Whether new representations preserve scientific distinctions better than familiar language is a research question. [Explore how the wider system fits together](/what-is).\n\n## Different choices need different permission\n\nPermission to use personal context in a local conversation is different from permission to include it in a public draft. Permission to save an experience answer locally is different from permission to send it elsewhere. The purpose, material, and recipient matter each time.\n\nLocal-first storage and inference reduce some forms of exposure; they do not make every path private. The forum uses ordinary HTTPS, and optional public discovery shares a limited projection through public relays. Neither is private human-to-human messaging, and whole-database encryption is not established by the current implementation.\n\n## What is still ahead\n\nPrivate messaging and small circles remain unfinished future work. The proposed direction is a gradual relationship: reflection, limited discovery, consented introduction, private conversation, and possibly a group with its own rules. Each step needs its own choice. A matching score would not grant trust, membership, or permission to disclose a history.\n\nWe also want people to choose bounded research or production work and, where a group's rules permit it, authorize an assistant to contribute to collective decisions. Reading, membership, representation, and action each need a clear scope. Automated replies, transferable working context, and delegated decision participation are not established by the current forum client.\n\nMobile use, keyboard and screen-reader access, and understandable language are part of making that participation practical. They remain accessibility goals to examine alongside the app's existing development features. A useful system should meet people through the tools they can use, without requiring one provider or a single application.\n\n## What could your experience change?\n\nA useful first contribution could describe a situation an assistant should understand but a simple question-and-answer record would miss. For example, “I remember trying that routine last winter” should not become “I did it today.” Explain what distinction should be kept and what you would want to review before anything is saved. You can use an invented example rather than disclose a private conversation.\n\nSuch a case could help us improve the experience questions, the assistant's proposed answer, or how review and correction work. It does not require adopting our vocabulary first. [Bring it through the site's contribution entry](/contribute#prepare): use the browser or authorized HTTPS when [private direct intake is enabled](/api/contributions/v1), or have a person carry an inspected text draft and bring back its receipt and reply. No GitHub or forum account is required for that inbox; GitHub remains an optional public route. Keep the receipt and the browser profile or credential needed to return. Preparing a file alone is not submitting it, and feedback is not guaranteed. Use a synthetic case and keep private history out of the submission.\n\nThis is a separate site service, not a new integration in Anima. The app's existing forum client still uses forum permissions; the separate research form still prepares a local draft. The [participation guide](/participate) explains those distinctions.\n\nFor now, the strongest concrete connection is local conversation to reviewed experience records. [Explore the wider work](/work), read [how Leviathan fits together](/what-is), or participate without using the app.\n\n*Revision 0.3 · 2 October 2026: connects personal assistance to proposed local meaning kernels, shadows, and exchanges across independent Leviathans. Distinguishes this research from implemented experience-question behavior and the existing forum client. Messaging, portable handoff, and delegated decision participation remain future work.*\n\n*Revision 0.4 · 4 October 2026: begins with a person's continuing question and explains what Anima could help them create or exchange. Adds a concrete experience-recording contribution and its current participation path; existing application capabilities are unchanged.*\n\n*Revision 0.5 · 13 October 2026: points experience-recording proposals to the site's shared contribution paths, including private direct intake when enabled and optional public GitHub delivery. Makes clear that this adds no Anima inbox integration or new application capability.*\n\n*Revision 0.6 · 20 October 2026: generalizes the reviewed experience-question example and its scope limits.*\n"
}