Build the system.Move closer to steel.

Every entry carries what works, what is synthetic, and what remains unsolved.

1. Vessel model

   VS-001 vessel and system model

   Synthetic model

   One synthetic concept vessel carried through every scene on this site — hull, structure, seven schematic build segments, and eleven illustrative layers split into physical systems and state layers.

   Why it matters to shipbuildingDownstream capabilities need shared object identity. When planning, material and inspection refer to different objects or segments, their state cannot be reconciled without resolving that disagreement.

   What actually works

   • One geometry drives geometry, system, dependency, production, inspection and delivery views
   • Layer isolation with dependency, owning roles, work packages, machine opportunities and inspection points
   • Keyboard-operable, reduced-motion-complete, screen-reader summarised

   What is synthetic

   • The hull, its dimensions and all identifiers are invented
   • System routes are illustrative, not engineered

   What it does not do

   • Not naval architecture: no hydrostatics, scantlings, stability or class compliance
   • Two-dimensional profile only — no 3D model, no volumes, no interference checking
   • Cannot be imported from or exported to any CAD or PLM system

   Unlocks nextA vessel model that can be populated from a real engineering source rather than authored.

   Vessel explorer
2. Build model

   Build Room — program to weld

   Synthetic model

   Ten levels of build hierarchy on VS-001, where every level answers the same six questions: what is this, what state, who owns it, what governs it, what is it waiting on, what does it contain.

   Why it matters to shipbuildingProduction is executed against work packages, not drawings. A system that cannot represent the package, its operations, its material and its inspection cannot tell anyone whether work can start.

   What actually works

   • Drill from program to a single weld operation and back to accepted configuration
   • Gates rendered explicitly: what each object is waiting on
   • Blocked state traced to a cause — a certificate mismatch holding two spools

   What is synthetic

   • The entire program, VS-P01, and every record inside it

   What it does not do

   • One authored program; no multi-vessel, no scheduling engine, no critical path
   • No write path — nothing can be updated, closed or approved
   • No integration with any planning or ERP system

   Unlocks nextA work-package engine that computes readiness instead of displaying an authored value.

   Build Room
3. Engineering control

   Governed answer core

   Working

   The layer Veldarium has actually earned: establishing which record governs a question under a given scope and date, keeping rejected evidence visible, and refusing to answer when authority cannot be established.

   Why it matters to shipbuildingIf a crew cannot determine which revision controls the work, the resulting decision can create rework, inspection, or material risk. This proof does not measure how often that occurs.

   What actually works

   • Authority, revision, supersession and applicability modelled explicitly rather than inferred from recency
   • Conflicting records surfaced instead of averaged into fluent prose
   • Evidence lineage attached to every answer, including what was rejected and why
   • Refusal path when no governing record can be established
   • Maker-checker human review for supported action types
   • Deterministic synthesis — no external language model in the public path

   What is synthetic

   • The public product proof applies the governed-answer contract to synthetic vessel VS-001 and work package WP-019

   What it does not do

   • Bounded to the document types listed in the capability registry
   • No production connector fleet — it does not read a live PLM, ERP or QMS
   • Scope and limitations are listed control by control on /trust, and they are real

   Unlocks nextApplying the same authority model to a bounded real engineering record set with an industry partner.

   Trust registry
4. Engineering control

   Change impact propagation

   Synthetic model

   CFG-029 evaluated across five canonical synthetic records: three known affected, one known not affected, and one where impact is not established.

   Why it matters to shipbuildingA useful impact record needs an evidenced boundary while preserving relations the available data cannot establish.

   What actually works

   • Three distinct states: known affected, known not affected, impact not established
   • The same canonical CFG-029 records feed the evaluator and the public impact surface
   • Illustrative profile positions are labelled as presentation coordinates, not engineering locations
   • Unknown impact is not silently promoted to not affected

   What is synthetic

   • The dependency graph is authored for VS-001, not derived from an engineering model

   What it does not do

   • Cannot discover dependencies — they are declared
   • No cost or schedule model behind the schedule-impact step
   • One scenario, not a general impact engine

   Unlocks nextDeriving the dependency graph from a real model so propagation is computed rather than authored.

   Change propagation
5. Build model

   Requirement to vessel

   Synthetic model

   One synthetic requirement followed down thirteen illustrative levels until it reaches a pipe-work operation, then changed to show which authored records would need review.

   Why it matters to shipbuildingVeldarium’s thesis is that making the path from requirement to physical work visible can make change review more disciplined; this surface does not measure an outcome.

   What actually works

   • Thirteen levels from requirement to vessel, each with an owner and a physical magnitude
   • A trace mode and an impact mode over the same ladder
   • Vessel isolation follows the walk — the ship lights up where the requirement lands

   What is synthetic

   • REQ-1847, its revision, and every object it touches

   What it does not do

   • One requirement, one path — not a requirements management system
   • No verification or traceability matrix
   • Magnitudes are illustrative, not calculated

   Unlocks nextMany requirements over one vessel, with coverage and orphan detection.

   Requirement to steel
6. Material

   Material readiness model

   Synthetic model

   Two items on the same nine-stage route into one work package. One clears. One is rejected at receipt verification on a certificate referencing a superseded revision — and the package is not ready.

   Why it matters to shipbuildingIf the material is not there, the work is not ready, and no planning system is entitled to say otherwise. Physical constraints have to govern the digital system rather than the reverse.

   What actually works

   • Readiness derived from kit completeness rather than asserted
   • A named cause for the hold, not a red badge
   • Material modelled as a first-class vessel layer

   What is synthetic

   • The items, the supplier, the certificate mismatch and the lead times

   What it does not do

   • No procurement, inventory or receiving integration
   • No supplier records, lead-time data or substitution workflow
   • Nine stages are illustrative of a real flow, not a configurable process

   Unlocks nextA material graph that reconciles a real BOM against real purchase and receipt state.

   Material flow
7. Production

   Program state view

   Synthetic model

   One operational picture of VS-001: engineering release, material readiness, structural completion, inspection closure, work-package counts, block state and named critical dependencies.

   Why it matters to shipbuildingThis teaching view asks what a responsible owner might need to inspect in one place; it makes no claim about how real programs currently review state.

   What actually works

   • Critical dependencies expressed as a constraint, a consequence and an owner
   • Authored synthetic block state uses the same identifiers as the drawing model; no roll-up is claimed

   What is synthetic

   • Every number on the view is invented for the synthetic program

   What it does not do

   • No ingestion, no history, no time series, no forecasting
   • Values do not move because nothing produces them
   • The value is the information architecture, not the figures

   Unlocks nextState computed from the build model instead of authored alongside it.

   Program state
8. Human + AI

   Human authority boundary

   Working

   A single published chain applied consistently across the whole site: machine recommends, system traces, human decides, configuration records, production executes, human accepts.

   Why it matters to shipbuildingAcceptance, safety and disposition carry legal and physical consequence. A system that blurs who decided something is not usable in a yard, however good its answers are.

   What actually works

   • Every operating stage states its AI role and its human gate separately
   • Each of the four actors carries an explicit boundary, including what AI does not do
   • Enforced by test: the deciding and accepting actor must be human

   What it does not do

   • This is an architectural commitment and a published position, not an enforcement mechanism
   • Nothing here prevents a future integration from violating it — that requires runtime controls

   Unlocks nextRuntime enforcement: an approval path that cannot be satisfied without a named person.

   System architecture
9. Production

   Industrial base dependency model

   Research

   The industrial base drawn as a four-tier load path from vessel to systems to fabrication to foundations, with the structural exposure at each node stated.

   Why it matters to shipbuildingA relevant constraint may sit in a yard, supplier, foundry, machining source, workforce pipeline, or another part of the industrial base; each claim needs facility-level evidence.

   What actually works

   • Structural argument published with per-node exposure
   • A facility-record standard published before any record exists

   What it does not do

   • The verified facility registry is empty. No record has been through the sourcing standard yet
   • No claim is made about any specific company
   • Source bodies are cited as publishers of data, not as partners

   Unlocks nextA first set of facility records that satisfy every required field, including a published coordinate.

   Industrial base
10. Human + AI

    Operational decision assembly

    Next

    Assemble live build state — work package, material, predecessors, governing revision, access — into a prepared decision a supervisor can accept or reject in one place.

    Why it matters to shipbuildingVeldarium’s hypothesis is that assembling the governing state may reduce decision-preparation burden. No current baseline, phone-call count, or time saving is claimed.

    What it does not do

    • Not started. The governed retrieval half exists; the operational assembly half does not
    • Requires the work-package engine to compute readiness first

    Unlocks nextThe first capability where Veldarium software changes what happens on a shift.

Every entry states its own limitsAn entry with nothing under “what actually works” has not been built