Look up terms as they arise in your reading. To try the protocol, follow
Your first draft; to implement it, consult the requirements
in the specification.
The project and website name, at legibleprotocol.com,
which publishes the specification, articles, books and reference code.
Participants create their spaces in their own environments.
LegibleProtocol is also the proposed GitHub organization name.
The specification's full name. Its edition identifies the formats and contracts
an implementation follows. Draft 0.4 is the specification edition presented
here; site and software versions are identified separately.
A space where participants can find, understand and develop one another's
contributions using records and their own tools. It may begin in a shared
folder or span projects and organizations. “Legibility space” describes
this condition of reading and collaboration.
The book's title and the practice it develops: making contributions and
relationships legible so possibilities for work and cooperation can form
before selection. The paper The Economics of Legibility presents the
economic argument.
Operating in Open Legibility means sustaining that path with participants,
tools and reading practices. It can take place in public, internal or combined
spaces. “Open” does not require publication: contribution can precede selection
within applicable access conditions. See the circuit.
An account of who formulated something, the criteria and choices they took responsibility for and the work they drew on. Provenance lets readers examine that participation alongside the result without needing the entire conversation that preceded it. It records what was declared; it does not automatically reconstruct someone's thought.
What someone adds to work: a question, observation, analysis, decision or solution, for example. The word includes contracted and paid work; it does not imply donating time. A contribution can exist before being selected for use.
Someone signing a record takes responsibility for the declaration it contains. This does not make them the author of every work it mentions. A third-party Trace, for example, identifies who declared the relationship while preserving authorship of the related work.
These are dimensions the book uses to examine work: defining what matters, evaluating alternatives against criteria and context, taking responsibility for the decision to act and carrying out what was chosen. One person can take part in all of them; the distinction assigns no positions and does not preclude AI assistance. When a test automates verification, it remains relevant to examine who defined what to test and which results would be sufficient.
A person and AI combining capabilities to work. Collaboration among pairs
also combines knowledge held by different people. Each participant retains
their authorship, including when results are incomplete or mistaken.
What a person seeks to accomplish, why it matters, which criteria they adopt
and which consequences they recognize. Records contextualize that participation,
including when execution is performed by AI. The expression fixes no universal
division of capabilities between humans and machines.
A declaration someone can take responsibility for and preserve without admission
by a central authority. Readers retain their judgment; sovereignty makes no
content true, imposes no recognition and grants no access to others' resources.
Signing, saving and sharing remain distinct operations.
A description of contributions participants choose to declare and their
relationships, with references to evidence. It may include a question that
changed an analysis, a correction or an adopted criterion. Conversation history
can serve as evidence; the genealogy highlights what each person claims
to have contributed.
Legible Draft is the draft profile's name. It uses the Genealogy 0.1
technical convention: Markdown with YAML metadata, including genealogy: "0.1".
A draft preserves a formulation, relationship or proposed recognition without
a cryptographic signature. The profile adds exchange and identification rules
while retaining the convention's fields.
A declaration presenting a formulation attributed to its author: a problem,
hypothesis, criterion or proposal, for example. A question can be a complete
formulation. An executed solution may be referenced as evidence.
A declaration of a specified relationship with another contribution or source,
such as building on, using, referencing or replicating it. The relationship's
declarant remains explicit; a third party may also assert it under the norm.
The relationship is an attributable claim for the reader to examine.
Retrospective recognition of a consequence received by the beneficiary.
It may be reputational, monetary or non-monetary under the specification.
For monetary recognition, the declared amount transferred and the description
of the benefit received express different things.
Formalization creates a new signed declaration after review of its content.
It may reference a draft's bytes; the draft remains informal. The signature
verifies integrity and attribution to a key. Saving the record privately and
sharing it remain separate choices.
Work, a version, document, test or other source referenced by a record.
Content may remain in its original environment under its own permissions.
The formulation can be the contribution itself; an external artifact is not required.
An accessible record can reference restricted evidence; the reference grants
neither access nor authorization for processing by an AI provider.
A draft's id is a local handle. Interchange also preserves origin and the
hash of exact bytes to distinguish replay, revision and conflict. A formal
record's id is computed from its canonical content. These identifiers serve
different purposes.
A private, optional reader perspective that can help express interests,
criteria and context when considering contributions. An implementation may
use it in reading and should identify its support. CH is neither an official
classification of people nor a network allocation rule.
The recognition convention adopted by a formalized Appoint, with an exact
version and digest. It defines recording recognition in the economic case
described, which may remain private. It distinguishes recognition from use
permission and the work's licence; it sets no amount or automatic payment
entitlement. The applicable text is NRC 0.2 in the specification.
An index helps locate records and can be rebuilt from the sources.
Contextual reading examines why a contribution may matter to someone's problem.
Tools, access and models determine what each environment offers.
A conversation may also bring out formulations that have not yet been recorded.
MCP is the interface through which an AI host can access configured
operations, such as reading drafts or preparing declarations. The reference
server identifies supported operations and editions. An Agent Skill is
an instruction package; loading it installs no model and authorizes no tools.
These are additional paths: first use through files can begin without them.
The paper's analytical names for two coordination orders: selection → preservation
→ search, in Regime A, and preservation → search → selection, in Regime B.
The latter allows contributions to remain available before decisions about use,
funding or hiring. Its economic advantage depends on costs and conditions in
each activity. Resources and allocation decisions remain scarce.
The convention by which a signatory adopts the declaration in the record in
good faith. This edition’s canonical text is Portuguese, with an English
reference translation. The signed content_schema identifies the commitment,
independently of the date. The signature does not certify truth, civil identity
or rights over evidence. See the specification.