Production blueprints

What was quoted is what is ordered: a production tier and a quality contract. Which floor actually runs the work is decided later, by routing. That separation is deliberate. It is what keeps the capacity configurable.

This is where the full package is required: the sound BOM, the drawings for the things that need them, and the suggested assembly, detailed below. Start with the published intake BOM schema, blank CSV template and EX-100 example to see how that input is structured. The thesis calls the BOM a strategic routing document: it determines not only what a product costs, but where it should be produced, which components dominate shipping cost, which suppliers must be close, and which production tier is economically justified.

Design-for-manufacturing analysis, tolerance stack-up, station design, fixture design, work instructions, quality gates: all compiled into a versioned blueprint and run against a certified hub's reference model first, catching collisions, takt-time gaps, and missing tooling while they are still cheap to fix. Like software: written once, reviewed, validated, then released. And written in a deliberate order, detailed below: the assembly definition first, the testing definition derived from it.

A sound structure is required, because every downstream decision reads it: levels and subassemblies that reflect how the product actually comes apart; part numbers and revisions for every line; quantities, materials, and finishes; make-or-buy flags; approved alternates; and long-lead or critical components marked as such.

Not every part needs a drawing pack, but specific things do: 2D drawings with tolerances for critical features, 3D models where geometry drives fixturing and robot access, interface and test specifications, and the quality requirements that will become inspection points.

The customer submits a proposed way to assemble the product: a sequence, an exploded view, whatever captures their intent. It is deliberately suggestive, never binding. The customer knows their product best; Moduloa's job is to know production. The suggestion enters as input, and industrialization turns it into a validated production flow: reviewed, risk-assessed, and released under change control. If the suggestion were binding, the blueprint and its validation discipline would collapse.

The customer's suggested sequence and the build tree go in; a production assembly definition comes out: stations in order, the fixture each step mounts, the parameters each step runs, and (read straight off the build tree) the exact points where a sub-assembly closes. Those closing points matter twice. They are where identity is minted and marked, and they are where access to everything inside is lost. So the definition records, step by step, what becomes reachable and what gets sealed away, because that access ledger is precisely what the testing definition consumes next.

Every test traces to a typed acceptance criterion from the intake BOM, the same criteria the costed BOM already priced. The testing definition adds the one thing intake cannot know: where each test can physically run. Each criterion is scheduled at the last moment its feature is still reachable, and whatever survives the final close becomes the end-of-line test. The result compiles into the blueprint as quality gates, and every result lands in the unit record.

Jig lead time is industrialization's critical path, and jigs must move at blueprint speed: a revision that changes a step often changes its fixture, and a flow under version control cannot wait out a supplier's quote cycle for its own tooling. So production aids live in the same discipline as everything else: an ID, a revision, a calibration status in the OS, released with the blueprint that needs them. The thesis names production-aid engineering (fixtures, jigs, test rigs, grippers, nests, gauges) as the first toolmaking capability worth rebuilding locally, and this is where that bet becomes operational.

Each jig has its own small build tree: a standard base or frame, machined or printed geometry derived from the customer's models, off-the-shelf hardware. It is industrialized by the same OS that industrializes the customer's product: same intake discipline, same revisions, same open economics. And the demand profile sizes the spec: 800 units a year, stable, earns printed nests on standard plates, not hardened steel. If demand grows into a tier upgrade, standardized interfaces mean the jigs are revised, not scrapped.

Humanoids do not remove the need for tooling; the thesis argues they raise its value, because the interface between robot, product, and process becomes a strategic asset. The roadmap hypothesis follows from that: hub humanoids fabricating and swapping the production aids themselves, printing a nest overnight, assembling a test rig from the standard frame catalogue. It is recorded as a hypothesis, exactly like the engraver bet on the marking page. Two claims in the register sit behind it, and they are not the same claim: P-13 is the capability bet, that toolmaking and fixture competence is recognized as a Norwegian industrial gap and rebuilt, while P-04 is a market bet, that robot tooling becomes a high-margin market for whoever sells it. Moduloa needs the capability either way. It does not need that market to be high-margin, and it will not price its own jigs as though it were in it.

None of this is a standard yet. It is a practice written down, and no hub has run it, because there are none. The Stage 0 timeline puts the tooling standard v0.1 in the 6 to 12 month step, and this section is the material it would be drawn from. Tier 3 of the published ladder is the modular cell, and the word doing the work there is modular, not robotic: a cell takes the next product when the fixturing bolts to the same interfaces, presents the part the same way twice, and carries a datasheet the OS is specified to read. Fixed cells are the reliable core of execution, and this standard is what would keep a fixed cell from being a single-product cell.

So v0.1 has to settle the base and frame interfaces, how part presentation is specified, the datasheet every aid carries (ID, revision, calibration interval, the model it derives from), which classes of aid are printed, machined or bought, and what a hub must hold to be audited against it. It would ship in the shape the intake standard already ships in: a schema, a blank template and a worked example, which the EX-100 set below already is. v1.0 is the version a certified hub is measured against, and that is Stage 1 work.

Confirm the BOM and part revisions, the referenced drawings and specifications, and the acceptance criteria. The assembly sequence and test access must agree with those inputs.

Check the hub's certified tier and required equipment, fixture revisions and calibration. Validate the flow against that hub's reference model for collisions, takt-time gaps and missing tooling. A released file alone does not establish portability.

The pilot must pass the quality gates at takt, with deviations closed into the blueprint or waived with a dated rationale. The version and unit records must remain traceable when the process is deployed.

EVT (verifying the design works at all) belongs to the customer, before intake; we do not design products. DVT's residue is what intake collects: the typed acceptance criteria are the pass limits the customer's validation produced. What this model owns is the PVT question: can the process build the product, at rate, inside the gates. That is the pilot, and it is why the intake standard insists on criteria, not intentions.

The pilot's minimum quantity is computed, not negotiated: new-to-hub processes, fixture count, tolerance stack, and the supplier confidence carried from quoting all raise or lower it. The same rule prices it for everyone under the published terms: one pilot is built into every engagement, and one pilot is guaranteed to be enough to reach production readiness. If it is not, the cost of being wrong is ours, and the post-pilot recalculation happens in the open.

The pilot exits when every gate passes at takt and every deviation is closed into the blueprint or waived with a dated rationale. The product record then carries the MP tag (mass-production ready), the flag routing treats as "this blueprint is production-real, place it anywhere its tier is certified." No tag, no routing; the tag is earned on the floor, not granted in a meeting.

Encoding process knowledge into a portable blueprint is the model's hardest claim: real factories hold tacit knowledge that resists being written down, and physical rollback is harder than software rollback. Whether a validated blueprint truly runs at a second certified hub without re-engineering is exactly what the register tracks: P-10 (validated production blueprints move between certified factories; portable production is commercially real). And the in-house jig commitment is a capability bet made in a country the thesis itself says lacks toolmaking depth. Rebuilding that layer starts with exactly the fixtures and rigs this page describes, and P-13 tracks whether it happens. See the register →

Use the thesis's production tiers to compare the intended trade-off between upfront engineering and portability. Follow the factory OS for versioning and deployment, and the economic model for unit pricing, capacity subscriptions and proposed blueprint deployment fees.