KNOBE PROTOCOL  ·  V1  ·  OPEN SPEC, FROZEN

Knowledge that keeps its context

A plain-text file that carries its source, history, limits, and obligations wherever it goes. No platform, no account, no server.

The problem

The words survive the trip. The context does not.

Watch what happens to a piece of work as it moves through ordinary systems.

Sealed at the source
attribution, consent, limits inside

LMS

copied into a course shell

CHAT

pasted into an AI conversation

DOC

summarized into a report

A bare quote
author gone, consent gone, limits gone

At every hop the words survive, and the conditions that made them responsible to use fall away. KNOBE's answer: make the context part of the file itself.

The mechanism

One plain-text file, three layers

A KNOBE is a single markdown file. A human reads the first two layers. A machine decodes the third. A verifier checks the hash.

course-notes.knobe.md

YAML frontmatter

title, author, date, license: scannable in two seconds

Markdown body

the work itself, readable in any editor, on any platform, forever

Sealed payload

attribution · parents · transformation history · fidelity limits · use conditions · AI policy · quarantine status

payload_hash = SHA-256 of the canonical form · the seal

Verify with one command

$ python3 lens.py course-notes.knobe.md
verified body: yes conformance: valid

The seal proves

  • the payload is byte-identical to what was sealed
  • the body matches its recorded hash
  • the context record is intact and machine-readable

It does not prove

  • that the content is true
  • that the attribution is honest
  • that consent was actually obtained

Integrity is the substrate. Interpretive context is the point.

The workflows

Five rooms, same file

Each domain uses different payload fields. Every workflow is the same move: seal the context in, verify it on arrival.

Education

  1. Instructor seals the assignment journey and posts it anywhere.
  2. Student writes and seals one file that cites the journey's hash.
  3. TA runs the cohort check; a regrade works from the sealed record, not memory.
attributionai_policyparents

The AI policy travels inside the submission, disclosed or prohibited, and the record outlives the semester.

Accessibility

  1. A work is adapted: captioned, simplified, translated.
  2. The adaptation binds itself to the original's hash and credits the adapter.
  3. Anyone can trace the lineage from adaptation back to source.
accessibilityadapted_from

Adaptations keep their provenance, and adapters keep their credit.

Research

  1. Participant seals the original; consent terms live in the payload.
  2. Researchers hold derivatives with lineage back to the sealed source.
  3. Consent travels with every derivative, not behind it in a drawer.
use_conditionsprivacy_level

Custody inverts: the participant keeps the original, and consent moves with the data.

Government

  1. A record is sealed at creation.
  2. A FOIA release derives from it, redactions recorded as transformations.
  3. The release names its parent; the derivation is checkable by anyone.
parentstransformations

A release that can prove which record it came from.

Enterprise & agents

  1. Sources are sealed with their use conditions.
  2. An agent asks permits() before acting: am I allowed to do this with this file?
  3. Syntheses declare their sources and inherit their obligations.
permitsquarantine_status

Agents that check the terms before they act, and get a deterministic answer.

The convergence

Every workflow asks the same question

What came with this?

Who made it, what it builds on, what it means to say, and what it was never meant to claim. KNOBE is informing infrastructure, not controlling infrastructure: it keeps that context attached and verifiable, in plain text, so that humans, institutions, and AI systems can understand what they are holding and use it responsibly.

A verified KNOBE is not necessarily trusted. A trusted KNOBE is not necessarily true. Context is how a reader, a classroom, or an agent tells the difference.

Back to knobe.org →