Status: ACTIVE
Class: POSITIONING_CONTROL
Scope: Public claims, repository README files, papers, demos, standards submissions, and governance-facing writing
Purpose: Lock the defensible claim surface. Prevent overclaim. Prevent underclaim through excessive hedging.
The field is not empty.
Runtime governance, delegation, audit trails, deterministic gates, and admissibility language are active areas of work. ALVIANTECH public material must not claim that nobody is building runtime agent governance or execution gating.
Permitted comparison shape:
No current product or public artefact appears to combine delegated authority, scope monotonicity, closed typed exits, cross-surface contradiction testing, sealed receipts, and a fail-closed execution boundary as one commit-boundary primitive.
This is a bounded comparison claim, not an exclusivity claim.
No unresolved interpretive state may bind directly into execution.
This is the spine.
It is intended as an executable invariant, not a philosophical claim.
ALVIANTECH builds runtime admissibility infrastructure for delegated AI systems, preventing unresolved interpretive states from binding into executable consequence.
The defensible ALVIANTECH surface is the combination of:
- Delegated authority — authority is traceable from a root human authority through a bounded chain.
- Scope monotonicity — delegated authority may narrow, but must not silently widen.
- Closed typed exits —
ALLOW,HOLD,DENY,SILENCEare controlled governance outcomes. - Cross-surface contradiction testing — active interpretive surfaces must remain structurally compatible before consequence is allowed.
- Sealed receipts — decisions produce inspectable evidence of what was attempted, checked, and returned.
- Fail-closed execution boundary — unresolved state blocks consequence.
The wedge is not generic agent governance.
The wedge is:
Cross-surface structural contradiction as a pre-execution admissibility gate.
At every delegation hop, the authority envelope of the delegate must be a subset of or equal to the authority envelope of the delegator.
No delegation hop may expand scope.
No action may proceed to execution if active interpretive surfaces produce mutually incompatible verdicts.
Cross-surface contradiction produces HOLD unless a surface returns DENY, in which case the result is DENY.
No unresolved interpretive state may bind directly into execution.
An unresolved state, where the contradiction engine cannot determine cross-surface compatibility, produces SILENCE.
SILENCE fails closed.
Do not claim:
- consciousness detection
- AGI
- emotional AI
- therapy or clinical application
- complete semantic understanding
- that nobody is building runtime governance
- that ALVIANTECH is the only governance primitive
- production readiness unless separately demonstrated
- certification, compliance, adoption, or standardisation unless independently evidenced
The following phrases are admissible when used with appropriate scope boundaries:
- bounded detection-and-prevention architecture at the pre-execution boundary
- runtime admissibility infrastructure for delegated AI systems
- typed pre-execution contradiction engine
- cross-surface structural contradiction as a pre-execution admissibility gate
- no current product appears to combine the six properties as one commit-boundary primitive
- runtime constitutional layer, as metaphor only
- admissibility substrate, with prior-art acknowledgement where relevant
- consequence-binding control system
Status: ACTIVE
Class: POSITIONING_CONTROL
Purpose: Prevent conceptual overload by assigning one primary mechanism to each public surface.
Conceptual overload collapses the spine.
No unresolved interpretive state may bind directly into execution.
Do not explain the entire lattice on any single surface.
Each surface earns the next one.
| Surface | Primary mechanism | Keep off |
|---|---|---|
| README | Delegated admissibility + typed exits | Full contradiction-engine theory; full receipt architecture; full delegation-chain formalism; standards positioning |
| Paper | Cross-surface contradiction engine | Product language; full receipt design; full delegation-chain implementation; marketing claims |
| Demo | Scope monotonicity + receipts | Coherence-surface theory; full positioning language; public-category claims; whole-system explanation |
| Standards submission | Formal invariants only | Product language; brand language; metaphor; narrative framing |
- Boundaries became inspectable.
- Frozen typed exits change category from application behaviour to governance primitive.
- Neighbouring work must be described accurately, not territorially.
Everything in the lattice is evidence for the spine.
The spine is not evidence for the lattice.
Do not use one surface to prove the entire system.
ALLOW, HOLD, DENY, SILENCE, EXIT_ENUM, SCOPE_MONOTONICITY, DecisionRecord, receipt, admissibility, commit boundary, execution boundary, and cross-surface are controlled terms.
No synonym substitution in public-facing governance material.
No runtime extension of EXIT_ENUM.
Frozen until formal human-authority revision.
Any public-facing document that departs from this register, including overclaim in marketing material or underclaim in standards submissions, must be flagged for governance review before publication.
Humour may decorate the control surface.
It may not mutate the governance rule.