Intuition Machine
Software

Two instruments, one kernel

One reads the words a system already runs and checks how far its claims sit from ground. The other stands at the front desk and decides which word runs next.

CM

Connection Machine

Witness graph and word tracing

  • Builds the fold-witness graph: which claims descend to evidence, and how far
  • Traces which of the twelve loops a system actually runs, and which fail to close
  • Detects letter deletions — the audit that compares reports to other reports
  • Flags orphaned conventions: standing rules with no living editor
IM

Intuition Machine

Typed dispatch

  • The selector made operational: routes arriving content to the handling loop
  • Types every route by level and column — no unindexed selection
  • Holds routing rules to the ceiling: quality-reading routes stay uncompiled
  • Keeps a live-judgment path and an alerting bypass permanently open
Connection Machine

What a witness graph is for

A convention-level claim is grounded exactly when a descent chain with provenance connects it to encounter-level records: the claim links to an interpretation, which links to a log, which links to something that actually happened. No such chain means no witness, and no witness means capture-vulnerable.

CM instruments that chain. It ingests the trace store, indexes the four strata of witness depth, and reports — per claim — whether a descent path to evidence exists and how many hops long it is. Because the test is mechanical, anyone can run it on any assertion.

GROUNDED — A DESCENT CHAIN WITH PROVENANCE the claim ν · convention interpretation record trace κ · execution log event what happened anyone can walk the chain and check UNGROUNDED — NO WITNESS, THEREFORE CAPTURE-VULNERABLE the claim ν · convention event what happened nothing to check the story against but the story
Fig. 07 — A claim with a descent chain, and one without

What it cannot do is eliminate the remainder. Convention is never fully groundable, so a narratable margin is permanent. CM's job is to make the groundable portion commonly computable, and to say plainly which claims fall outside it.

Intuition Machine

Why dispatch needs a guard

A strange complaint arrives on Monday morning. It might be an operations matter, an audit trigger, or early market intelligence. Every loop is healthy; the diagnostic layers examine each one, find nothing, and return nothing — because the failure was that the item reached the wrong loop.

IM makes the selector explicit. Every route is typed by level and column. Rules are marked as compiled or live, and the mix is matched to stage — compiled selection is appropriate standardisation at commodity stage and lethal at genesis.

EVERY LOOP HEALTHY — AND THE CONTENT REACHED THE WRONG ONE arriving item a strange complaint Sel the front desk typed · regimed · owned ⦿O handled as operations MISROUTE — competent, and wrong ⦿I early intelligence where it belonged ⦿R reflexive quality ROUTE STARVATION — never engaged The diagnostic layers examine each loop, find all of them healthy, and return nothing.
Fig. 08 — Misroute and route starvation at the front desk

Two guards are structural rather than configurable. Routes that must read quality-level content cannot be fully compiled, so no rulebook becomes the sole gate on quality-sensing. And every compiled rule set is itself a standing convention, so it carries a witness and an owner — because orphaned routing rules are the dead hand at the front desk.

Status

What is claimed, and what is not

The kernel these instruments implement is verified: seven axioms, nine load-bearing theorems, and 43 of 43 checks passing in the canonical model. The constructs the dispatch layer rests on are newer and carry weaker labels — the selector is FRAMEWORK with its formalisation deferred to an open problem, and the ceiling conjecture is SEMI-FORMAL pending a model family not yet built.

Those labels are printed here for the same reason they are printed in the specification. A tool that reports on grounding should be honest about its own.

Build it, or find out which layer is missing.

Design engagements build the architecture for a specific organisation. Diagnostic engagements establish which of its loops actually close, which of its claims reach evidence, and which of its standing rules no longer have an owner.