The first edition, prepared for launch
These articles are prepared from reviewed pre-release source. Publishing these articles does not announce a public app release. Research posts link dated evidence rather than claiming live experimental status.
Subscribe by RSS. Release announcements will appear only after an actual approved release; until then, see the pre-release development summary.
Development and research
Research ·
Completion claims and verifier integrity in coding agents
The working manuscript audits overlapping success/subversion signals, turn-limit interpretation and historical-binary confounds. Read the 13-page PDF and its evidence limits.
Development ·
A green check is only as strong as its gate
How to read Commandem verification without mistaking a completed turn for a correct result.
Research ·
Reporting an experiment without laundering its failures
Why controls, delivery evidence, infrastructure categories and null results belong in the same report.
Development ·
Continuity is an identity problem before it is a memory feature
Separate history visibility, native resume and reviewed cross-provider context before calling a conversation resumable.
Development ·
Local-first orchestration does not mean local-only AI
Where Commandem keeps control and evidence, and where provider and community services remain separate trust boundaries.
Editorial standards
Every entry states its date, scope and reviewed source. We distinguish working behavior from planned capabilities, separate experimental populations and retain negative results. Corrections should be visible; material post-observation amendments should not silently replace earlier criteria.
Authors are identified as Commandem maintainers. Articles must not export private account identifiers, transcripts, secrets or unpublished study artifacts. Provider-policy claims require current authoritative sources.
Release announcements
No approved public release is announced in this edition. An announcement needs a versioned artifact, platform and verification details, migration guidance and known limitations.