Counters are real and start at zero. They move when the floor does. Last reviewed 2026-08-01.
Interface artifact · Computed 2026-09-16
Twenty-two interfaces have a value on both sides. Three agree.
In a public community dataset covering one lunar process, both the providing and the receiving organisation stated a value for the same interface twenty-two times. The two values were compared as stated.
- 22
- Interfaces with a value stated on both sides
- 3
- Where the two values agree
- 19
- Where they do not
- 6
- Where the receiving system needs more than the providing system plans to deliver
All three power interfaces disagree.
Power is the case with the least room for a difference of opinion, because the consumer either draws what it needs or stops working. It is also the case where every matched pair in this dataset disagrees.
| Interface | What the matched pair shows |
|---|---|
| Power interface 1 | The provider plans to deliver 40 percent of what the consumer states it will draw. |
| Power interface 2 | The stated values disagree. The magnitude is held in the computation and is not published here. |
| Power interface 3 | The stated values disagree. The magnitude is held in the computation and is not published here. |
Method: every interface in the dataset where both sides stated a value was matched pair by pair, and the two values compared as written, with no normalisation, no unit conversion beyond the one the dataset already used, and no judgement about which side is right.
The contributing organisations are anonymised. Where this artifact needs to distinguish them it uses Contributor A, B and C.
Twenty-two matched pairs from seven organisations describing one process indicate that interface disagreement is measurable and worth measuring at scale. This is an indication, not a proof. Quantifying the problem across the lunar architecture requires a larger supporting count before this page can report its scale.
Four turns, each one paid for by a customer.
Measure
Every article FORGE qualifies is measured against acceptance criteria on calibrated, traceable instruments. The customer gets the data package; the corpus keeps the measurement.
Accumulate
Measurements are retained in a structured corpus: component behavior under vacuum at temperature, screening outcomes, instrument drift, uncertainty actuals. Data nobody else is positioned to collect, because nobody sells this measurement to U.S. defense customers today.
Compound
Each campaign makes the next one better: tighter uncertainty budgets, faster test design, screening criteria backed by observed populations instead of handbook values.
Return
The corpus flows back out as faster qualification cycles, priced screening data, and eventually reference datasets the industry checks against. The lab is how we acquire it; the corpus is what appreciates.
What a corpus record carries.
The shape is fixed before the first record exists, because a corpus is only as good as the discipline of its first entry. Contents accumulate as the floor runs; the shape is the commitment.
- article
- Part identity, lot, and configuration under test
- campaign
- Test campaign, procedure revision, and acceptance criteria
- environment
- Vacuum level, temperature profile, and fixture configuration
- measurement
- Measured values with units and timestamps
- uncertainty
- Stated uncertainty budget for the measurement chain
- traceability
- Calibration chain for every instrument in the setup
- disposition
- Pass, fail, or nonconformance reference
The capability behind the corpus.
The corpus exists because FORGE is being built to sell the measurement nobody else in the U.S. offers commercially. See the qualification gap map for the four-way comparison, or bring us the part you need proven.