Knowledge answerable to life

No person can follow every development that might change their life. Depending entirely on one institution or system to interpret it would leave too much outside our own questions and judgment. We want useful production, understanding, and meaningful participation to develop together: people and communities should be able to make things, examine their basis, and shape the questions and decisions that follow.

Leviathan is the system we are developing toward that purpose. Its central language research asks how observations, concepts, methods, and values can connect across people, AI systems, and independent communities. A contribution might change the question itself, open a new field of inquiry, or make a better tool possible. What a thought changes in life, whom it helps, and whom it overlooks belong within the inquiry.

Learning from another place

Consider the fictional Delta care cooperative and Reed methods collective. Delta notices a difference in an animal's sensor recordings. Reed contributes a comparison method and suggests another technique from a different area of signal processing. The suggestion could create a new route to investigation; it is not yet a demonstrated explanation of the animal's condition.

The contribution changes their question: which signal changes can the comparison distinguish from device effects? The groups retain different data policies. When Reed finds a missing device condition in its original method, Delta can find the interpretation that used it and reconsider that interpretation. The observations remain available, and the group's term for a recording change need not change merely because its conclusion changes.

Together they draft a synthetic comparison plan, varying recording settings with a fixed input and then changing the input with fixed settings. It would compare the corrected method with Reed's candidate for false flags and missed changes; it reports no experimental outcome. Each group decides its own next step.

This design example expresses both invention and correction. It asks how a useful connection can cross a boundary, how its limits can travel with it, and how later experience can improve the work. Those relationships are what we want to develop beyond one example.

A changing relationship with knowledge

Language carries experience across generations. Instruments extend observation beyond unaided senses. Models offer further ways to interpret information and generate things we can examine. The forms in which knowledge becomes usable may change as those methods develop.

Human language carries care and curiosity alongside prejudice and conflict. A category can reveal something about a life or obscure what matters about it. Credo, the philosophical vision informing Leviathan, invites us to treat our descriptions as revisable encounters with a world whose possible distinctions we have not exhausted.

We want to investigate representations beyond familiar words and categories while keeping their connections to observations and consequences open to examination. Evidence and values have different roles: a measurement bears on a claim, while a commitment helps explain which consequences matter. Values also influence the questions and methods we choose. Research is not thereby free of every value or assumption.

Whitenote's aspiration to make science and thought independent of human emotion remains part of the personal starting point. This vision's current working distinction is to keep wishes from substituting for evidence while allowing care, lived experience, and values to shape questions and face criticism. The relationship between these positions remains an open philosophical question.

We use ethical guardianship of language to mean continuing responsibility for these relationships, including our own descriptions. Better concepts should help us understand and care more accurately; greater expressive or technical power should not make its holder the unquestionable interpreter of everyone else.

Many Levis, many Leviathans

A Levi is a participating unit with a continuing context. It may be connected to a person, group, value, animal, place, or natural process. Cooperating structures can form Leviathans, including relationships among other Leviathans. Independent structures may overlap, separate, and work at different levels of detail. There is no required universal root.

Each could maintain a local meaning kernel connecting the concepts, commitments, relationships, and revision practices it actually uses. A group might learn a measurement technique without adopting its author's ethics. Several animal ethics communities could disagree while contributing observations to a shared question.

Independence also needs examination in practice. Participants may depend on the same models, funding, instruments, or sources. A declared identity does not establish a separate perspective, and a connection that carries information does not automatically transfer authority.

A language capable of changing

The first authored language foundation and structured catalog give these relationships definitions, reasons, and revision conditions. The ambition is a language through which an observation can inform a question, a question can motivate a method, and a result can change later work.

Such a language may combine human explanation, structured records, images, sensor streams, executable methods, and representations learned by machines. Its design should make consequential differences expressible without requiring every future distinction to fit today's vocabulary. A new representation must be examined for what it preserves, loses, and makes possible.

Accepting a record as a reference for a stated participant and purpose is what we mean here by canonization. It is separate from formatting or summarizing that record. Acceptance does not make a proposition universally true. Translation should expose unresolved differences and the context a summary leaves behind.

Terms, principles, rules, and procedures for revising them need versions because their meaning and use can change. Earlier meanings remain part of the history. New observations may require a new concept; a corrected inference may leave the concept intact. Reasons for the change should determine which happened.

Discovery and learning that improve their own methods

A shadow is our working name for a doubt, missing distinction, contradiction, or lesson retained for further inquiry. A later observation, sensor, or idea might give us a way to pursue it. Recording uncertainty should make exploration possible, rather than merely catalogue what is unresolved.

Creativity belongs here. A method from an apparently unrelated field may suggest a useful experiment or artifact. Its result might change how another participant investigates a problem. Eventually, a system could propose improvements to the representations and learning procedures through which these discoveries happen.

System-level learning can involve a retained working practice that improves later use of methods on unfamiliar problems, even when model weights stay unchanged. Revising or storing a procedure alone does not demonstrate that improvement. Learned representations and model training are further possibilities with their own evaluation needs.

Recursive self-improvement and possible paths toward artificial general intelligence are substantial research ambitions. They require evidence that an improvement helps later work on unfamiliar problems, at an intelligible cost. Changing an evaluator to obtain a better score would not establish that result. Representation and self-improvement remain open questions.

Discovery, transmission, selective learning, and the endurance or revision of commitments are related achievements. A success in one does not establish them all. The five foundational questions expose these distinctions and invite objections, including whether simpler existing tools could meet the same needs.

Let our own work change

In the research programme and its dated local inquiry, a reviewer noticed that the delivered context omitted the exact decision it was asked to assess. We supplied the decision and revised the procedure for later requests. The original gap and response remain visible. This was a concrete procedural correction in coordinated development work. Improved subsequent use on unfamiliar work has not yet been demonstrated.

That inquiry now records exact uses, review returns, local agreements, decision adoption, and application separately. These starting pieces make some of our reasoning inspectable. The wider independent network, learned language, and autonomous discovery methods remain to be developed and evaluated.

Sources should support the particular claims they bear on. A philosophical argument can clarify a commitment; a study can test a mechanism. Neither proves the entire architecture. The structure must remain able to change when its own experiments, participants, or consequences expose a mistake.

Care and participation

Animal welfare is an intended first field of application. Our ethical starting point is that greater capacity to understand and affect other lives brings responsibility to attend to them. Their interests matter even when they cannot contest our account in human language.

Tasma explores observation, Otomat practical care, and Anima personal assistance and participation. Other independent projects may offer different methods and better solutions. A sensor reading needs interpretation, and representing an animal's interests requires more than naming an advocate.

The same purpose extends to research, software, and environmental work. People should be able to contribute through what they care about, including a question unrelated to our existing projects. Privacy and permission matter when sharing; support should have a stated use and should not buy a preferred conclusion.

The current language and optional profiles are proposals open to scoped adoption. Forum categories, inherited values, and operating arrangements can be deliberately redesigned. Their earlier forms remain history, not a permanent common constitution. Existing software permissions still apply until changed through their own implementation.

Read how the parts connect, bring the Markdown introduction to an assistant you choose, or contribute directly through the available paths. Mimar's Whitenote preserves the personal starting point. What Leviathan becomes should remain open to what others help us understand.

Revision 0.7 · 3 October 2026: connects the vision to the freshly authored language foundation and structured catalog. Opens the vocabulary, domain profiles, and forum design to revision from their purpose while preserving historical records. Distinguishes authorship, scoped adoption, implementation, and actual permissions.

Revision 0.8 · 4 October 2026: leads with purpose and a fictional exchange before introducing terms. Develops creative discovery alongside correction, distinguishes concept revision from changed inference, and connects the ambition to the actual local inquiry.

Revision 0.9 · 4 October 2026: brings production, understanding, and participation together through language research. Gives the fictional exchange a proposed comparison plan, clarifies system-level learning, and states a current distinction between evidence and values while leaving its relationship to Whitenote's aspiration open. The historical text remains unchanged.