Evidence · Status as of 21 October 2026

What we claim, and what we do not.

Every number on this site is an internal result, and each one names its population. Some earlier records did not deserve the trust they asked for, so we withdrew them. Both facts belong on the same page.

RC3 freeze677 / 677 checks · internal
V2 integrated tests410 / 410
Soak recordsWithdrawn 30 Aug 2026
External auditAudit not completed
01 How to read this page

Three populations, never added together.

410 integrated tests belong to the COSRYX V2 mission recorder and gate G1. 692 product tests belong to the twelve products and the Hybrid Moat. 677 RC3 security-gate checks belong to the internal security freeze. They test different things, and adding them would produce a number that means nothing.

All three are internal: run by DCS on DCS-controlled infrastructure. Internal results are not an external audit, not certification and not flight qualification.

Simulated results are labelled "Simulated", every time.

02 What is claimed

Internal results, with their names on.

410/410COSRYX V2 integrated tests; code gates C0–C9 all GOG1 complete
5,447of 5,447 manifest entries reproduced exactlyG0 complete
677/677RC3 internal security-gate checksFreeze complete
692/692product tests across 12 products and the Hybrid Moat, 0 failInternal
SIM-LUNAR-01

System Study 001 · 100/100

Lunar scenario checks across the product line.

Simulated

MARS · MCC + EDGE

Conformance 46 / 0

46 conformance checks passed, 0 failed, under Mars delay.

Simulated

MARS · INTEGRITY

Radiation-integrity simulation 18 / 18

A simulation of bit upsets, not a radiation test.

Simulated

03 The RC3 freeze

An internal security freeze, complete.

RC3 is the third release candidate of the COSRYX mission-evidence stack, frozen on 6 August 2026 after 677 of 677 internal security-gate checks passed.

Freezing means the code under review stops moving, so an external auditor examines exactly what we tested. The external audit has been initiated and is not completed.

We do not publish release archives, audit packs or evidence bundles on this site.

04 Product tests by group

How 692 is made.

GroupProduct testsStatus
Foundation: PQ-Agility, DSRP, AI Sovereignty Oracle249 · 0 CVEsTested Simulation-validated
S1–S7: LENS, WATCH, WEATHER, MCC, EDGE, NEXUS, SWARM TRUST258Built Simulation-validated
N13 SPACETRACE, S8 SSA ACCOUNTABILITY, S9 MFG PROVENANCE96Built Ground-demonstrated
Hybrid Moat: HYBRID-01, 02, 0389Built Definition-of-done sign-off pending
Total692Internal

Per-product counts are on the products page.

05 Withdrawal notice · 30 August 2026

Earlier soak records are withdrawn.

Earlier internal soak records were withdrawn on 30 August 2026 because the verdict logic could not report a failure; corrected records will be re-issued as a set. As of 21 October 2026 they remain withdrawn, pending that re-issue. No figure from them appears anywhere on this site.

Why all six, not one. The defect was shared. In each run, real measurements passed through verdict logic that could not reach a failing outcome: a gate reading a field nothing writes, a timer that expired before the run began, a check that did not exist. Withdrawing one document while leaving the others standing would publish a set that disagreed with itself, so the whole set was withdrawn together and will be re-issued together.

No corrected figure, deliberately. Publishing corrected numbers one document at a time is how a set comes to contradict itself. Corrected figures will be released only as a coordinated set, once the runs have been re-gated against controls that can report a failure.

What was not wrong. The reconciliation found no fabrication at the source: the individual figures came from machine-generated closing records. An earlier generation of the documents had held the numbers back with explicit pending markers.

The finding, in one line: the measurements were largely sound; the judgements placed on top of them were not. That distinction is why the original documents are preserved unchanged as the historical record, rather than deleted.

06 Document by document

Six documents, six findings.

Summarised from the 30 August notice. We describe the defect in each, and no result.

24-HOUR HYBRID SOAK

Receipt coverage was never checked

The record issued fewer receipts than it ran cycles, and none of the verdict checks read receipt coverage, so the gap could not affect the verdict however large it grew.

Withdrawn

7-DAY SOAK

A verifier that always said yes

The run was configured with a receipt verifier that returned true unconditionally. The cycles ran; nothing about them was verified.

Withdrawn

30-DAY SOAK

A timer that fired at once

The configured duration exceeded the largest delay the timer accepts. On overflow it fired within milliseconds and the run marked itself passed.

Withdrawn

72-HOUR TRIPLE-HYBRID SOAK

Cycle completeness was invisible

The run completed materially fewer cycles than configured, but no verdict function read completeness. An uptime threshold was declared and read by nothing.

Withdrawn

15-DAY SOAK

"No error" over a recorded error

The run fell short of its schedule, and a no-error check passed while the record's own last-error field was not empty.

Withdrawn

PORTFOLIO SUMMARY

Sums of withdrawn numbers

It repeated the withdrawn results and added aggregate figures that appear in no closing record. This is the one place where the numbers were not read from a record.

Withdrawn

07 Scope

What this evidence proves, and what it does not.

Proves

  • The code exists and passes its own internal tests, by the counts above
  • COSRYX V2 is code-complete (G0 and G1)
  • The internal security freeze RC3 is complete
  • Named scenarios behaved as specified in simulation

Does not prove

  • Flight qualification or flight readiness
  • Radiation testing (it is next)
  • External audit or certification of any kind
  • Customer deployment or agency acceptance
  • Anything from the withdrawn soak records
08 Required statements

Printed in full.

Internal security freeze (RC3) complete · External audit not completed · No flight qualification claimed.
The post-quantum (PQ) leg uses ML-DSA-65 (FIPS 204) via the R-Series engine. Key custody in this stage is software; hardware custody certification is pending. Not flight-qualified. Not independently audited.
No launch contract is signed.
09 How a claim changes

No status moves without evidence.

A status on this site changes only when three things exist: updated evidence, a green test suite, and sign-off by the named owner. A status never moves up because a page needs a better headline.

When a claim turns out to be wrong, it comes down, the reason is published, and the correction is dated. The soak withdrawal above is the standing example.

The full status vocabulary and its definitions are on the governance page.

10 Questions

About the evidence.

Can I see the withdrawn documents?

The originals are preserved unchanged as the historical record. They are not presented as current evidence, and we do not repeat their figures here.

When will corrected soak records be published?

When the runs have been re-gated against controls that can report a failure, and the set can be re-issued together. We will announce it in the newsroom.

Does the withdrawal affect the 410, 692 or 677 counts?

No. Those are separate test populations with their own verdict logic. The withdrawal concerns the soak records only.

Who is doing the external audit?

The process has been initiated. We do not name firms before an engagement is confirmed, and the audit is not completed.

11 Scrutiny welcome

Found something we got wrong?

Tell us. Corrections are published, dated and kept visible.

ContactNewsroom