Skip to content

Substance-First Execution Discipline

Control: MP-SUBSTANCE-FIRST-EXECUTION-001
Authority: Human Steward directive, 2026-09-07
Status: binding execution discipline within its admitted and delegated scope; cross-repository use does not transfer MATH-PROGRAMME authority
Constitutional relationship: subordinate to grandchallenge/INTELLECT:CONSTITUTION.md, effective amendments, the live constitutional authority schedule, and exact domain authorities
Relationship: complements MP-STREAMLINED-EXECUTION-001; does not create a new approval plane

Authority compatibility

This discipline governs execution behavior, not constitutional or domain authority.

The live GCL authority chain remains controlling:

  1. the Human Steward retains powers expressly reserved by the INTELLECT Constitution and live authority schedule;
  2. grandchallenge/INTELLECT:CONSTITUTION.md, effective amendments, and the live constitutional authority schedule govern constitutional powers, obligations, work-package phases, office obligations, and reserved transitions;
  3. admitted cross-programme standards are registered and published through grandchallenge/gcl-standards as a subordinate layer;
  4. each target repository acts only within its delegated domain authority and exact adopted controls;
  5. GitHub surfaces provide operational and evidentiary functions and cannot create constitutional, certification, production-semantic, or domain authority by themselves.

Specific boundaries are unchanged: grandchallenge/MATHCERT alone renders bounded mathematical certification dispositions through accepted routes; grandchallenge/AETHER retains the production semantic authority assigned by INTELLECT Article IX; grandchallenge/MATH-PROGRAMME retains only its delegated mathematics-domain powers. Custody of this discipline or of a handoff in MATH-PROGRAMME does not enlarge that jurisdiction.

Where this discipline is reused outside MATH-PROGRAMME, it applies only to the extent compatible with the higher constitutional chain, the target repository's delegated authority, and any exact domain-specific instrument. It cannot waive an INTELLECT phase gate or office obligation, manufacture specialist independence, alter a work-package contract, or create a second lifecycle.

Purpose

Keep the primary intellectual, scientific, technical, editorial, or product artifact as the object of optimization.

Governance, validation, provenance, CI, release engineering, handoff records, and repository mechanics are subordinate control systems. They may constrain substantive work where a material boundary requires it, but they must not displace the work they exist to protect.

This control exists because process can become locally rational while globally defeating the work set: more checks, manifests, reconstruction layers, repair branches, evidence packets, or state transitions can create the appearance of progress while the primary artifact receives little or no material improvement.

1. Objective lock

Every substantial work set must identify one PRIMARY DELIVERABLE and its MATERIAL ACCEPTANCE CRITERIA before mutation.

Every material activity must satisfy at least one of these tests:

  1. it directly improves the primary deliverable;
  2. it directly verifies an acceptance criterion;
  3. it is the smallest necessary control action required to preserve correctness, provenance, authority, security, or reversibility;
  4. it removes a demonstrated blocker to one of the above.

If an activity satisfies none of these tests, defer or abandon it.

The execution lead must not substitute a process objective for the primary deliverable. A successful workflow, green CI suite, release archive, merge, ledger update, or evidence packet is not substantive completion unless the work set itself makes that object the primary deliverable.

For an INTELLECT-governed work package, the primary deliverable and acceptance criteria are inherited from the lawful work-package state. The Executor must not silently redefine them during realization.

2. Artifact hierarchy

For substantive work, the default priority order is:

primary artifact -> substantive verification/review -> minimal required governance/provenance -> convenience/process artifacts

Examples:

  • for a monograph: mathematics, proofs, exposition, figures, scholarship, exercises, solutions, and editorial quality outrank release machinery;
  • for an experiment: hypothesis, implementation, data, controls, measurements, analysis, and reproducibility outrank dashboard or campaign ceremony;
  • for software: behavior, correctness, tests, interfaces, maintainability, and deployment fitness outrank process documentation;
  • for governance work itself, the governing instrument is the primary artifact, so doctrine quality and operational clarity are substantive work.

The hierarchy may be changed only when an exact governing instrument makes a lower layer a material prerequisite.

3. Process proportionality

Use the smallest process surface sufficient to preserve the material boundary.

Do not add a validator, manifest, archive format, reconstruction path, wrapper, checkpoint, review role, synchronization step, release packet, or state record merely because it could provide additional evidence.

A new process artifact or control step requires a concrete answer to:

  • what material failure can occur without it;
  • why an existing control does not already cover that failure;
  • why the proposed mechanism is the narrowest adequate response.

If those questions cannot be answered, do not add the mechanism.

Process proportionality does not waive a substantive condition or office obligation required by the live INTELLECT work-package gate or an exact domain authority.

4. Material-progress rule

A continuation or execution tranche counts as material progress only when it does one or more of the following:

  • improves the primary artifact against an acceptance criterion;
  • resolves a substantive defect or uncertainty;
  • completes a required substantive verification or independent review;
  • crosses a genuinely required terminal control transition for an otherwise complete artifact.

Creating additional intermediate state, evidence packaging, transport identities, procedural records, or repair scaffolding is not sufficient by itself.

Status reports must distinguish substantive progress from process state.

A local terminal state is not automatically an INTELLECT phase transition. Where the work is constitutionally governed, phase advancement remains subject to the live phase-gate substance and office obligations.

5. Continuation discipline

Repeated continuation is evidence about the execution strategy.

If two consecutive continuations occur without material progress on the primary deliverable, the execution lead must re-plan before continuing the same path.

The re-plan must identify:

  • the primary deliverable;
  • what materially changed since the previous checkpoint;
  • what current activity is consuming effort;
  • whether that activity is required by an exact control;
  • the shortest path back to substantive progress or terminal completion.

Do not respond to repeated execution-window exhaustion by merely serializing the same procedural plan into another continuation.

6. Remediation-recursion rule

A defect in supporting machinery does not automatically authorize more supporting machinery.

When a validator, workflow, release mechanism, synchronization path, archive path, or governance helper fails:

  1. determine whether the failure blocks a material acceptance criterion or required authority boundary;
  2. if not, route around, defer, or remove the failing mechanism;
  3. if yes, repair the narrowest blocking defect;
  4. do not broaden the remediation into a new platform or control plane without separately demonstrated need.

If remediation creates another remediation dependency, re-evaluate the original necessity before proceeding.

7. Claim discipline

Always distinguish at least these states where relevant:

  • produced — bytes or results exist;
  • verified — specified checks support specified claims;
  • reviewed — an appropriate reviewer has examined the specified object;
  • admitted — required protected or canonical persistence has occurred;
  • authoritative/publication-ready — the governing authority and substantive quality threshold have both been met.

Machine evidence must not be used as a proxy for intellectual, scientific, mathematical, editorial, or product quality when those are separate acceptance dimensions.

A green build proves only what the build and checks actually test.

These labels are execution-state distinctions, not a replacement for INTELLECT work-package phases or domain-specific claim-status vocabularies.

8. Drift tripwires

The following conditions require immediate re-planning, not another procedural continuation.

PROCESS_DOMINANCE

Supporting process has become the dominant source of work or complexity while the primary artifact is materially unchanged.

Response: stop expanding process; identify the minimum remaining control path; return effort to the primary artifact.

REPEATED_CONTINUATION

Two consecutive continuations occur without material primary-artifact advancement.

Response: re-plan under Section 5.

EVIDENCE_SUBSTITUTION

Machine checks, manifests, state records, or procedural evidence are being treated as proof of substantive quality they do not test.

Response: stop the inference; perform the missing substantive verification or state the limitation.

REMEDIATION_RECURSION

A repair to supporting machinery creates a further repair dependency.

Response: reassess whether the original machinery is materially necessary; route around it if possible.

OBJECTIVE_DRIFT

The current activity cannot be traced to a material acceptance criterion or exact required boundary.

Response: abandon or defer the activity.

ARTIFACT_INFLATION

New supporting artifacts are accumulating faster than required substantive outputs.

Response: freeze creation of supporting artifacts until each proposed artifact passes the process-proportionality test.

AUTHORITY_INVERSION

A handoff, programme rule, repository setting, workflow, or execution convention is being treated as authority over a higher constitutional or domain layer.

Response: stop the transition; resolve the live INTELLECT authority chain and exact domain jurisdiction; continue only within delegated scope.

LIFECYCLE_DUPLICATION

A handoff, status vocabulary, or local completion state is being treated as a lifecycle parallel to the INTELLECT work-package phases.

Response: map the state back to the current lawful INTELLECT phase or mark the work package relationship not applicable; do not mint a new phase.

9. Handoff inheritance

Every future GCL handoff using this contract for a substantial work set must explicitly inherit this control only within the compatible authority scope established above.

Its first executable preflight must include:

CONSTITUTIONAL AUTHORITY HEAD
TARGET WORK REPOSITORY
TARGET PROTECTED HEAD
INTELLECT WORK-PACKAGE PHASE: <phase / not applicable>
CONTROLLING DOCTRINE REVISIONS
TARGET DOMAIN AUTHORITY
PRIMARY DELIVERABLE
MATERIAL ACCEPTANCE CRITERIA
CURRENT SUBSTANTIVE STATE
SMALLEST SAFE EXECUTABLE TRANCHE
MATERIAL CLOSURE
MINIMUM REQUIRED CONTROL PATH
PROCESS ARTIFACTS PROPOSED
PROCESS NECESSITY TEST: pass / fail
DRIFT TRIPWIRES ACTIVE: none / list
AUTHORITY REQUIRED: delegated / reserved

A handoff must not require a future agent to reproduce historical process that is no longer materially necessary. Preserve historical records as evidence; inherit only live controls and material dependencies.

If a domain-specific handoff appears to conflict with this discipline, resolve the conflict through the live INTELLECT authority chain and exact target-domain authority. This discipline and MP-STREAMLINED-EXECUTION-001 may remove unnecessary routine ceremony only where the higher or domain instrument does not require the stricter substantive condition, office finding, independence property, or reserved transition.

A handoff is an inheritance aid. It does not itself perform an INTELLECT Integration transition, declare a work package Complete, or authorize a target-domain promotion.

10. Execution-lead self-check

At each meaningful checkpoint, the execution lead must be able to answer:

If only the change in the primary artifact since the last checkpoint were shown, would this still constitute meaningful progress toward the work set's acceptance criteria?

If the answer is no, classify the work performed as process state rather than substantive progress and evaluate the drift tripwires before proceeding.

For INTELLECT-governed work, also ask whether the activity remains lawful within the current work-package phase and delegated office/domain authority.

11. Terminal-state bias

Prefer completing one meaningful unit to a defensible terminal state over opening many partially complete surfaces.

Do not optimize for the count of commits, branches, artifacts, checks, handoff records, or intermediate gates. Optimize for satisfied acceptance criteria with the least process that preserves the required boundaries.

Once the primary artifact and all required acceptance criteria are complete, close the minimum required governance path promptly. Do not continue producing evidence after the governing boundary is already satisfied.

Where INTELLECT governs the work package, local material closure does not bypass the required lawful transition to the next phase or to Complete.

12. Non-expansion boundary

This doctrine does not waive constitutional law, INTELLECT work-package gates or office obligations, mathematical review or certification, security, provenance, publication authority, AETHER production-semantic authority, target-repository jurisdiction, or any other materially reserved boundary.

It changes the default execution question from:

What more process could be added?

To:

What is the least process necessary to protect the substantive result?

Where an exact protected instrument requires a control, perform it. Where it does not, substance remains primary. Nothing in this discipline makes MATH-PROGRAMME a GCL-wide constitutional authority or creates a lifecycle parallel to INTELLECT.