# ClearAI positioning

## What it is

ClearAI is an **ontology discovery and exploration platform, native to DSH**: AI grows a living **domain ontology** inside real projects — the project's own domain language (vocabulary), the entries established through the loop, and their graphs. The Epistemic Loop is the ontology's **production process**: form a hypothesis, inspect or test it, record observations, evaluate the result; every term is admitted with a basis, every entry is established by evidence, and refuted claims stay on record as negative knowledge.

The dividing line in one sentence: other knowledge graphs pile up edges by extraction and assertion; here **every edge has to be earned through the loop**. The direction of truth never changes — world → evidence → fact → grown into the ontology — so **the ontology adjudicates nothing; it only houses what has been adjudicated**.

ClearAI is not an autonomous oracle, a general-purpose chat wrapper, or a claim of recursive self-improvement (RSI). It helps a person reason better; it does not replace judgment.

## Audience

ClearAI is for curious, technically minded people working on questions where sources conflict, evidence is incomplete, or the path to an answer is not yet clear: researchers, engineers, analysts, builders, and independent investigators.

## Differentiation

Most AI products optimize for fluent answers, task completion, or conversation history. ClearAI is organized around the **status of knowledge**: what is hypothesized, what was checked, what was observed, what remains uncertain, and why a conclusion should be trusted. DSH provides the native harness: durable workspaces, explicit execution modes, governed plans, evidence records, and reversible history.

## Four promises

1. **Make uncertainty explicit.** Separate questions, hypotheses, observations, evaluations, evidence, and facts.
2. **Make inquiry traceable.** Keep the path from prompt to work to conclusion inspectable and recoverable.
3. **Make conclusions appropriately bounded.** Preserve confidence, provenance, unresolved alternatives, and the conditions under which a claim holds.
4. **Give knowledge a form.** Established conclusions settle into ontology entries (vocabulary + content + graphs), retrievable next round by concept instead of by re-reading prose.

## Boundaries

- ClearAI supports inquiry; it does not guarantee truth or make high-stakes decisions for the user.
- It can search, analyze, test, and organize evidence, but it cannot turn missing evidence into certainty.
- Automation is bounded by tools, workspace permissions, source quality, and explicit verification.
- The user remains responsible for judgment, approvals, interpretation, and consequential action.
- ClearAI does not claim RSI, consciousness, independent agency, or unrestricted autonomy.

## Terminology

- **Epistemic harness:** The surrounding system that structures inquiry, records its state, and keeps claims tied to evidence.
- **Epistemic Loop:** The repeatable cycle of hypothesis, verification, observation, evaluation, evidence, and bounded conclusion.
- **Harness:** The DSH-native execution and governance layer around the model and workspace—not a claim about the model itself.
- **Evidence:** A recorded support for a claim, with provenance and limits.
- **Fact:** A claim accepted for the current workspace under its recorded evidence and scope; bounded by its scope. A fact is the ontology's **content unit** — written as an assertion in the ontology's vocabulary, shelved at the moment of promotion.
- **Ontology:** The whole of a project knowledge base's language and content: vocabulary (convention), established entries (experience), and their graphs. In product contexts it means the domain ontology by default; the process ontology is the plugin's internal flow and is not included.
- **Ontology shelf:** The shelf view of the ontology's established content: every entry carries its boundary and support level; older, unstructured entries stay on the shelf, marked "unstructured".
- **Bounded conclusion:** A conclusion stated with its confidence, conditions, alternatives, and remaining uncertainty.