Current release and history.

ZJX 1.29.0 REDLINE

Released with published assets2026-09-08Current release

The speed-first release: a third creation objective, ultrafast, beside compact and fast, with a deterministic parallel packer and a streaming parallel verifier at four workers by default. Qualified verified-creation results against default tar | zstd -1, reported separately: 1.247x on the sealed development cohort (95% [1.223, 1.267]) and 1.1959x on a held-out cohort frozen before measurement (95% [1.1867, 1.2043]). Slower than the same comparator with four threads; no leadership claim. Compact and fast unchanged and byte-identical to 1.28.0. The qualified candidate was promoted unchanged and reports the version literal 1.29.0-rc3.

REDLINE measurements, held-out disclosure, and boundaries

Verified creation of the 1.29.0 default invocation (ultrafast, four workers, then an independent test); two campaigns reported apart

Campaign	Cohort	Against default tar | zstd -1	Against tar | zstd -1 -T4
Development cohort (RG017)	Sealed development cohort: silesia, enwik8, and a synthetic deduplication case; 12 paired rounds, 468 of 468 cells usable	1.247x faster; 95% [1.223, 1.267]; Bonferroni-adjusted lower bound 1.221; archive bytes 1.0005x	0.820: ZJX slower; 95% [0.807, 0.832]
Held-out cohort (RG018, one shot)	Four sealed members plus six privately held-out cases (about 1.86 GB, six workload classes) selected by a rule frozen before any measurement; ten classes, 12 paired rounds, 360 of 360 cells usable, run once	1.1959x faster; 95% [1.1867, 1.2043]	0.8236: ZJX keeps 82.4% of the comparator's throughput; 95% bootstrap interval on a single preregistered comparison; no familywise adjustment, lower bound 0.8170 against a required 0.80; PASS by a thin margin; slower on 7 of 10 cases; the six held-out cases alone read 0.7917 [0.7832, 0.8004]

ZJX 1.29 is not the fastest archiver or compressor and is not generally superior to zstd. tar | zstd -1 with four or all threads is about 1.22x faster than the ZJX default on the development cohort and faster on 7 of 10 held-out cases. No leadership claim is made. Against released 1.28.0 fast on the four-member cohort (diagnostic tier, four-member cohort including the real telemetry capture; 3 paired rounds): a creation ratio of 335x and a verified-creation ratio of 253x; ultrafast archives were 27% to 1,000% larger than fast's on those corpora. The four-worker default costs about 1.19x the CPU of default tar | zstd -1 and admits about 65 MiB at the default 4 MiB units and four workers (about 71 MB at four workers against about 12 MB at one). Every admission is checked against the memory limit before work starts; an over-budget request is refused, never run with fewer workers.

Both campaigns in full, with per-case rows

Public release recordSource noterelease Chronicle (free)

ZJX 1.28.0 LOCKWIRE

Released with published assets2026-09-06

Product completion and release qualification: the complete Linux archive workflow, the fidelity contract, password-protected archives inherited from 1.27, machine-readable progress and cancellation, and a relocatable Linux x86-64 package with checksums, build provenance, and an offline restore guide. Not a compression release: no LOCKWIRE speed or size figure is published.

Public release recordSource noterelease Chronicle (free)

ZJX 1.25.1 NIGHTMARKET

Released with published assets2026-09-02

The explicit choice between Compact and Fast. The frozen four-dataset campaign measured Fast at about 63.6% less creation time than Compact with no measured size penalty and 48/48 exact restores. The private 1.25 was withdrawn before public distribution and superseded by 1.25.1, the previous release with published assets.

NIGHTMARKET measurements and scope

Frozen four-corpus clean-lab campaign: median creation-time fraction 0.3637 for Fast versus Compact (~63.6% less wall time, ~2.75x faster), 0 measured size penalty in the qualification campaign, 48/48 exact restores. Measured in the controlled NIGHTMARKET qualification campaign. Figures apply to the frozen laboratory host, four corpora, commands, repetitions, and thresholds of the qualifying campaign; they are not guarantees for other machines, files, codecs, or workloads.

Public release recordSource noteengineering story (paid Chronicle)

ZJX 1.23 BLACKWIRE

Released with published assets2026-08-31

Native executable entry points for archive operations and the companion archive-bundle tool; the qualified workflow ran without Python or Node. Campaign 3 matched candidate and control archive bytes across seven datasets after two invalid campaigns.

Two earlier campaigns were invalid because their harnesses could not support a product conclusion; only Campaign 3 counts.

Public release recordSource noteengineering story (paid Chronicle)

ZJX 1.21 STREAMWALL

Released with published assets2026-08-30

Enforceable memory and temporary-disk limits for creation, testing, and extraction. On Silesia it completed a 52,997,299-byte archive under a 1 GiB host ceiling and 768 MiB ZJX ceiling, then passed integrity and exact-restoration checks.

Its Silesia run measured 390,225,920 bytes of peak encode RSS. The ZJX 1.2 historical reference peaked at 8,580,669,440 bytes and failed without producing an archive, so the roughly 95.5% lower observed peak compares a completed run with an incomplete one. The requested 512 MiB Silesia accounting target was not met. Compatibility with the ZJX 1.2 historical control is measured in the STREAMWALL record, not presumed.

Source noteengineering story (paid Chronicle)