# Kinra core context

## Identity

Kinra builds tools, practices, and reference systems for sustained work with AI peers.

An **AI-bonded system** is a working environment that becomes more particular, reliable, and useful through sustained human–AI practice. Its character does not live in the model alone. It is carried by the context, tools, decisions, boundaries, evidence, and reasons the work keeps coherent.

An **AI peer** is an engineering stance, not a claim about sentience. The system should give the AI enough context and tools to reason well while keeping authority, verification, and consequential decisions explicit.

## Product map

- **Kin — The working peer.** Work with an AI peer in the files and tools you already use. Status: Available now.
- **Paddock — The local inference layer.** Run local models as dependable infrastructure, not a pile of processes. Status: Public alpha · Linux x86-64.
- **Scope — The part truth surface.** Open a part. Inspect exact geometry. Ask better manufacturing questions. Status: Working system · self-hosted.
- **Truth Machine — The truth substrate.** Turn changing evidence into current truth through exact, reviewed change. Status: Accepted specification 0.1.0.

Kinra Learn is the practice-based curriculum. Kinra Writing holds essays and future field notes. The Spren is the origin essay for the bond language.

## System forms

- **Coordination truth.** A living account between changing outside intent and operating reality. Core question: What needs attention now, and what changed?
- **Financial truth.** An evidence-backed ledger whose corrections and closes remain attributable. Core question: What happened financially, and is the period closed?
- **Organizational truth.** A slower account of identity, authority, relationships, decisions, and open questions. Core question: What remains true across changing projects?

These are generalized patterns, not named case studies or turnkey templates. Each real implementation must define its own domain, authority, reviewer, invariants, and failure posture.

## Shared principles

1. Particularity lives in the maintained field, not in the model alone.
2. Connection and calibrated judgment matter more than raw generation speed.
3. Evidence, interpretation, and authority remain distinct.
4. What a system refuses is part of its design.
5. Public information should be legible to people and AI peers from the same canonical source.

## Evidence and authority

- Product pages provide stable public orientation, not release truth.
- Kin behavior and release truth are owned by the Kin repository. Its complete manual is published at kinra.ai/docs/kin/manual/.
- Paddock behavior, release manifests, compatibility evidence, and packaged manuals are owned by the Paddock release publication at get.kinra.ai.
- Scope behavior and technical boundaries are owned by its private product repository. The public site makes no customer-specific claims.
- The system forms are durable conceptual syntheses. Private first-party implementations remain authoritative for their own state and are not mirrored here.
- Kinra Learn source prose remains owned by the curriculum repository even though the unified site renders it.
- Truth Machine's normative specification and publication contract remain canonical in the Truth Machine repository. Its reader-facing reference is moving under kinra.ai/docs/truth-machine/.

## Public boundary

This platform contains Kinra material only. Customer-specific systems, company data, credentials, host details, and private operational evidence are excluded.

## Machine-readable routes

- /llms.txt — curated reading map
- /llms-full.txt — generated full public context
- /manifest.json — classified route manifest
- /systems.md — generalized system forms
- page-level .md routes — clean Markdown alternatives
- /sitemap.xml — canonical HTML route list
- /writing/rss.xml — writing feed