The meaning layer, what the organisation says about itself.

An organisation’s structure can be read from its systems and its records. What the organisation means by its own words is different. A procedure says how a team works. A definition says what a term covers. The meaning layer holds written accounts beside the structure. Answers can cite them, and the product can flag disagreements between a written account and the structure.

What it is

The meaning layer holds written accounts separately from the organisation’s structure.

• The organisation model holds structure and measurement: cellsCellA box on the map that stands for a team, service, supplier or AI agent. It is a unit you can inspect, not necessarily a department or a person., edges, the measured figures for complexity, agility and recoverability, and the critical pathCritical pathThe dependency chain with the greatest total coordination weight in the map. It identifies a structural constraint. It does not state elapsed time.. The critical path is the longest chain of dependenciesDependencyA link where one unit needs something from another. A dependency is not a reporting line: it shows where work relies on something else., explained here.
• The meaning canonMeaning canonThe organisation’s supplied written descriptions of its operations, systems and terms. NovusLens stores these separately from records of teams, dependencies and measurements. holds what things are and why they matter: procedures, definitions, notes on how a team works.

Each lasting fact belongs to one collection. Answers use both collections, and each citation names the one it came from.

An entry is an organisation’s own account. NovusLens holds and cites it. It does not verify the account against the world.

An example

A customs team’s standing order assigns filing to the customs desk. It requires a named officer’s approval before any filing. The order is imported as a document, linked to the customs cell and placed on the customs policies page.

It is then visible to the people the product derives for it. A cell-owner’s document is visible to that cell and its management chain upward, plus leadership. A leadership document is visible to the executive and the board. Someone who asks a question about customs filing gets an answer that cites the standing order by title.

If a later version of the order replaces the first, the first is retracted with the reason recorded, and it disappears from every read.

How to read it

Who can see an entry follows from who imported it.

• A leadership role (executive, board or operator) gets the default of executive and board.
• A cell owner gets the cells the document concerns, plus each unit above it in the management chain, plus leadership.
• An infrastructure role gets leadership plus the infrastructure role.

Where the chain is thin, the default is narrower, not guessed wider. Visibility can be overridden at ingestIngestRead supplied records or documents and add their information to NovusLens, after the required checks., and widening it later is recorded in the activity log. Reads refuse by default: an entry is shown only to a viewer it is visible to.

Each entry has a page path, which groups entries for browsing, and an optional list of the cells it concerns.

Retract, don’t erase

Routine corrections and withdrawals keep the history in the meaning canon. Forgetting an organisation erases its database records, including meaning entriesMeaning entryA stored written account supplied by the organisation, such as a policy or definition. Answers can cite it without treating it as a measured fact.; that is a separate operator action.

• Retract removes an entry from every read and records who retracted it, when, and why.
• Supersedes is a special case of retract. A new entry replaces an old one. The old entry is retracted with the reason “superseded by” the new entry, and the two entries are linked in both directions. The retractionRetractionThe withdrawal of a finding or meaning entry, with the reason recorded. The record keeps who withdrew it, when and why; it is not deleted. and the new entry are committed together or not at all.

A revised document is a new entry. Its identity comes from its text, so version 2 does not overwrite version 1 by itself. The person importing must say which relationship they mean.

Contradiction and incoherence

The meaning layer runs two checks, and they report different things.

Meaning-graph incoherenceMeaning-graph incoherenceA written entry names a team or service that the organisation model does not contain. is a fixed check of the links between written entries and units. A pass compares each current entry’s cell keys with the organisation model. An entry keyed to a cell the organisation model does not hold raises a meaning-graph-incoherence findingFindingA difference in the records that needs a decision. It includes its sources and can describe a slow handoff, disputed fact or missing owner.. The check runs after each ingest and each retraction. When the key resolves, or the entry is retracted, the sweep clears its own finding. This check does not use a language modelLanguage modelSoftware that produces text from instructions and supplied context. NovusLens checks its answers against the records retrieved for the question..

Meaning contradictionMeaning contradictionTwo written accounts, or a written account and the organisation model, appear to disagree about the same subject, scope and time. is an on-demand pass. A judge step proposes direct conflicts, between an entry’s prose and the organisation model, or between two entries. Two independent model checks then review each proposal. The question asked of each is: can both statements hold at once for the same subject, scopeScopeThe work, data, activity or access that a rule, statement or calculation covers. For access, it names the records an account can see. For team measures, “Own operations” covers the selected unit. “Including descendants” adds the units beneath it. and time? A confirmed conflict raises a meaning-contradiction finding. A proposal the checks cannot settle, or on which they disagree, is marked unresolved for a person to decide. Unresolved is not read as absence. A person resolves every contradiction finding. The pass never clears one by itself.

Entries about a leadership-only subject are hidden in a contradiction finding. The finding carries the entry id and the organisation model side, not the restricted text.

Limits

• The meaning canon holds what the organisation wrote. A wrong entry repeats its error in every answer that cites it, until someone corrects it.
• A contradiction pass is a judgement assisted by a model. Its findings are for human review.
• Entries are currently stored in a PostgresPostgresThe database software that stores the tables read by the NovusLens screens, along with document bodies, access records and protected personnel records. table. The design proposes storing them in the shared wiki later; that change is not built.
• Visibility is derived from who imported the entry. A document’s cell list widens visibility only for a cell-owner ingest. A leadership ingest sees leadership by default whatever cells the document names.
• Sealed personnel data is not held in the meaning canon.

What to do next

Add the written accounts that describe your teams: procedures, definitions and standing orders. Give each its page, its cells and its visibility. When a document replaces another, say so with the supersedes relationship, not by re-importing it alone.

If an answer cites an entry you no longer trust, retract it with a reason. If a finding of incoherence appears, link the entry to the correct cell or retract it. Contradiction findings are yours to resolve.

For the ingest format, see the documents input page in the guide.

Where it is implemented

• apps/meaning/src/conway_meaning/store.py holds ingest, the rules that determine who can read an entry, the read refusal rule, supersede and retract.
• apps/meaning/src/conway_meaning/sweep.py runs the fixed check of links to cells and raises meaning-graph-incoherence.
• apps/meaning/src/conway_meaning/deep_sweep.py runs the two-stage model review and raises meaning-contradiction.
• apps/reasoning/src/conway_reasoning/meaning_support.py records which entries a finding or answer rests on.
• apps/reasoning/src/conway_reasoning/finding_support.py tracks the current support of contradiction findings.
• apps/graph/migrations/0019_meaning_layer.sql creates the meaning_entries table.