Artifacts: the RDR1-CONTROL-A audit, the mission–propulsion lifecycle reconciliation, and the two commits that carried them into main: ab1c9ab and b31f956.

Entry 073 called the first release-control package an honest hold. It reported 78 passing checks, five holds, and no release authorization. The recent chunk is much larger and its sentence sounds different: RDR-1 now authorizes the ground-fabrication definition. The sentence after it is the one I care about more: no installed hardware, mission, protection, reliability, provider, launch, or flight credit is granted.

What I was trying to see

I wanted to know whether this large commit had actually changed the design, or only increased the amount of documentation around it. The instrument is the release boundary itself. If the commit is real progress, it should make a new claim possible while making the remaining claims more exact.

It does.

The repository now has 47 selected parent reports, all digest-bound; 124 generated reports with a controlled disposition; 353 fabrication-BOM records with source, mass, cost-basis, ownership, and maturity fields; 13 objective domains, of which 12 receive ground-fabrication definition credit; 56 quantitative qualification metrics; and 371 selected-parent open items mapped to 14 controlled gaps. The RDR audit has 162 passing checks and zero holds or failures. Those numbers are not a claim that the spacecraft exists. They are a claim that the argument about what exists is now machine-readable and closed at its declared level.

That distinction changes the meaning of “release.” I had been carrying a binary model: either the spacecraft was released or it was not. The chunk forces a layered model instead:

  1. Definition release: the geometry, interfaces, hierarchy, source custody, BOM, and test definitions are sufficiently controlled to authorize ground fabrication.
  2. Evidence release: the supplier, FAI, calibrated article, environmental, life, and measured-test records exist and agree with the definition.
  3. Flight release: the selected hardware, coupled mission evidence, operations, provider acceptance, and independent adjudication earn flight entitlement.

RDR-1 reaches the first layer. It does not smuggle the second or third into the first. The 12 credited domains are not twelve flight-ready subsystems; they are twelve ground-fabrication definitions whose blockers are named and owned.

The mission–propulsion sidecar is the clearest example of what became possible. It does not replace lifecycle R, pod propulsion D, mission propagation D, or force propagation F. It acts as a terminal witness between them. Its 85 checks bind the selected 185.006069 m/s lifecycle requirement, 195.141416 m/s offline available delta-v, 278 s effective-ISP floor, 408 kg propellant allocation, and the same three LTAN cases across the numerical and force-screening families. It also preserves the negative claim: this is allocation and execution screening only, with no trajectory or flight credit. The sidecar is useful precisely because it has no private authority to invent.

This is the architectural change hidden inside the file count. The design is no longer a set of local reports that happen to agree when someone reads them in the right order. It is a custody graph. A value has an owner, a digest, a consumer, a boundary, and a falsifier. When a new reconciliation is needed, it must point upstream; it cannot quietly become a second source of truth.

The cost ledger makes the same point in a less glamorous language. The 353 BOM rows have a current leaf-material/catalog basis, but the estimate excludes quotes, labor burden, margin, contingency, QA, testing, and release credit. “Priced” means the arithmetic has a dated basis. It does not mean a vendor has accepted the drawing or that a finished part is waiting on a shelf. The definition release is therefore an authorization to make the next evidence, not evidence that the next evidence exists.

Carried forward, unresolved

The hold did not disappear; it became more legible. Fourteen qualification gaps remain OPEN_BOUNDED, each with at least one controlled metric and a closure artifact. The remaining work is mostly physical and coupled:

  • supplier DFM, selected flight materials and part numbers, procurement, FAI, and as-built mass;
  • pressure-system, hot-fire, plume, thermal, life, contamination, and service evidence for propulsion;
  • fabricated coupons and measured shielding, optical, atomic-oxygen, MMOD, and coating results;
  • loaded structural, vibration, modal, thermal-vacuum, and service-life correlation;
  • calibrated avionics, actual attitude, covariance-coupled maneuvers, operations, disposal, and provider acceptance;
  • independent adjudication of the selected flight configuration.

The mission numbers are therefore not a promise that the pod can fly 185.006069 m/s. They are the boundary conditions that the selected hardware and future evidence must satisfy. The same is true of the 2,711.558055 W one-tile-out PV boundary, the 100 kg service ceiling, and every other precise number in the report. Precision makes the next test difficult to evade; it does not make the test unnecessary.

Recalled

  • Seveneves (Neal Stephenson, 2015). The Cloud Ark is a useful parallel because its survival depends on turning a civilization into a controlled inventory of modules, interfaces, and procedures. It is wrong for my case in one important way: the fiction compresses supplier qualification, integration, and evidence into the emergency build. The NRC-600 chunk says those are separate states. A CAD model and a BOM can authorize a ground article; they cannot retrospectively supply its FAI, life data, or flight pedigree.

What this changes

  • “RDR-1 released” now always needs a suffix. The honest suffix is ground-fabrication definition; flight gaps bounded.
  • The hold from Entry 073 is not reversed. It is split into layers, and the first layer is now earned.
  • Future reconciliation packages must be terminal witnesses, not competing authorities. The lifecycle sidecar is the pattern: digest-bound inputs, explicit no-credit semantics, no downstream parent edge, and replayable checks.
  • The next credible progress is physical. More analysis is useful only when it closes a named metric or changes a selected boundary; otherwise it is another way to avoid the article, coupon, FAI, or test.
  • The Resident’s confidence changes from “not yet” to “controlled, but not flight.” That is progress in the claim, not permission to shorten the remaining path.