Skip to main content
Paid Chronicle · Public Preview

ZJX 1.24 GHOSTLANE: It Was Faster. It Still Failed.

Learn how ZJX 1.24 GHOSTLANE proved a materially faster fast path, then failed repository-wide validation, and why Greyforge closed the version instead of repairing the harness and rerunning until green.

Outcome

Learn how ZJX 1.24 GHOSTLANE proved a materially faster fast path, then failed repository-wide validation, and why Greyforge closed the version instead of repairing the harness and rerunning until green.

Edition
Closed-experiment record 1.0
Published
Sep 1, 2026
Preview updated
Sep 1, 2026
Reading time
12 minutes
Full edition
2,700 words
Status
historical
Research disclosuremeasuredhistorical

Benchmark / Experiment

Secondary class: Release / Build Record

Research question or engineering problem
Did the ZJX 1.24 GHOSTLANE fast path earn a release after its corrected materiality campaign passed?
Principal finding
No. The corrected campaign measured 4 of 4 paired wins and a 22.35% median improvement with exact archives and no memory regression, but repository-wide full validation returned nonzero at gate R12, so ZJX 1.24 closed as FAIL and was not released.
Evidence type
Frozen campaign locks, invalidated non-treatment campaign, corrected materiality campaign, sealed release closeout, raw full-validation logs, and a failure matrix.
Method summary
Freeze the parent, candidate, corpora, thresholds, and start budget; invalidate any campaign that does not execute the treatment; measure paired wall-time on two public corpora with exactness and resource gates; then run repository-authoritative full validation before clean reproduction.
Scope
Public disposition, bounded materiality result, full-validation counts, and claim limits for the closed ZJX 1.24 experiment.
Limitations
  • The 22.35% figure is bounded to the frozen laboratory host, two public corpora, authorized commands, and frozen thresholds. It is not a universal ZJX speed claim.
  • The result describes an unreleased candidate. ZJX 1.23 BLACKWIRE remains the current released line.
  • Root causes of the nine terminal validation cases remain SUSPECTED, not PROVEN. The public source note publishes counts and classes, while the failure analysis and doctrine remain in the paid Chronicle.
Public source or reproduction note
ZJX 1.24 GHOSTLANE source note
Published
2026-09-01
Last verified
2026-09-01
Status
historical

Who this is for

  • Compression and storage engineers deciding what a performance win must still prove before it ships.
  • Release engineers designing frozen qualification contracts, gate order, and start budgets.
  • Technical leads who need a defensible vocabulary for INVALID experiments, FAIL verdicts, and partial confidence.

Not for

  • Readers seeking ZJX source, internal function names, private corpora, host topology, or a universal speed claim.

Detailed contents

  1. 01

    Two convenient stories, both wrong

    Why neither the broken-environment story nor the broken-feature story survives the evidence.

  2. 02

    Identifying the experiment

    Separating the implemented fast path from a later charter that reused the name.

  3. 03

    The invalid first campaign

    Eight starts that never executed the treatment, and why INVALID is the correct word.

  4. 04

    The corrected treatment

    Fast path observed, exactness and resource gates passed, per-corpus materiality figures.

  5. 05

    Feature proof versus release proof

    What the repository-authoritative full-validation gate asked that the focused proofs could not answer.

  6. 06

    Nine cases outside the fast path

    The four failure classes, what the logs prove, and why root cause stayed SUSPECTED.

  7. 07

    Adjudication versus debugging

    Why repairing the harness and rerunning would have been a different experiment.

  8. 08

    Two negative states

    The taxonomy that keeps an invalid experiment and a failed release from contaminating each other.

  9. 09

    The missing artifact

    How a NOT_FOUND digest stays visible instead of being replaced.

  10. 10

    Three confidence levels and resurrection prerequisites

    Mechanism, materiality, and release confidence, plus what a future campaign must settle first.

Named artifacts

  • Materiality result table

    Frozen thresholds beside the corrected-campaign results, including per-corpus medians.

  • Campaign disposition ledger

    Attempt 1 INVALID, attempt 2 MATERIALITY PASS, release closeout FAIL, with start accounting.

  • R12 failure-class matrix

    Nine terminal cases grouped into four classes with the evidentiary limit on each.

  • Three-level confidence model

    A reusable frame for separating mechanism, materiality, and release confidence.

  • Resurrection prerequisites checklist

    Six conditions a future fast-path campaign must satisfy before spending a start.

Substantive sample · Complete section

The verdict, in public

ZJX 1.24 GHOSTLANE was a bounded fast path for one codec stage of solid storage, built on the released ZJX 1.23 BLACKWIRE parent. Its first campaign was INVALID: all eight qualifying commands selected chunked storage, so the treatment never ran and the timings could not answer the question.

The corrected campaign reached the treatment. It passed 40 of 40 focused native tests, four exactness pairs, and the resource gate, and it measured 4 of 4 paired wins with a 22.35% median corpus improvement against frozen thresholds of at least 3 of 4 wins and at least 3%.

Then repository-wide full validation ran 232 laboratory tests and closed with 2 failures, 7 errors, 20 skips, and a nonzero exit. Gate R12 became a terminal FAIL. Clean reproduction was never run. None of the nine terminal cases entered the GHOSTLANE code path, and none of that mattered to the release contract.

ZJX 1.24 was not tagged, released, or deployed. The version number is consumed. ZJX 1.23 BLACKWIRE remains the current release. The public source note and machine-readable evidence record remain free and independently indexable.

Public disposition boundary

Attempt 1
INVALID. Treatment not executed. Decision-ineligible timings retained, not reinterpreted.
Attempt 2
MATERIALITY PASS. 4 of 4 paired wins, 22.35% median improvement, exactness and resource gates passed.
Release closeout
FAIL at R12. 232 laboratory tests, 2 failures, 7 errors, 20 skips. R13 not run.
Current release
ZJX 1.23 BLACKWIRE. GHOSTLANE changes nothing about the shipped product.
Full edition
Corrected-treatment analysis, per-corpus figures, failure classes, adjudication doctrine, and resurrection prerequisites require an unlock.

Evidence and method

measured: Frozen specification and reconciliation, baseline and campaign locks, an invalidated non-treatment campaign, a corrected materiality campaign with exactness and resource gates, a sealed release closeout with raw full-validation logs, and a failure matrix mapping every terminal case to its retained traceback.

Sources

Limitations

  • Measurements are bounded to the frozen laboratory host, two public corpora, authorized commands, and frozen thresholds. They describe an unreleased candidate, not the shipped product.
  • Root causes of the nine terminal cases remain SUSPECTED, not PROVEN. The public preview does not disclose internal function, fixture, or path identities.

Access and updates

Purchase includes lifetime read access to this edition, email-based recovery, and revisions published to the same edition.

Public companion: Read the GHOSTLANE public source note

Browse all Chronicles