← Back to journal

Make the Evidence Portable, Not the Private Environment

What a useful technical handoff should preserve without leaking the machine that produced it.

Evidence basisKalu, Korpus, and frozen Ushman source/docs snapshot reviewed September 17, 2026; no new runtime qualification performed
DisclosureEngineering or methodology note based on a September 17, 2026 snapshot. Source-level behavior is not a public availability or universal compatibility claim. Any scenarios are illustrative, not customer outcomes.
Related to Kalu KodeExplore Kalu Kode →

Scope: Engineering or methodology note based on a September 17, 2026 snapshot. Source-level behavior is not a public availability or universal compatibility claim. Any scenarios are illustrative, not customer outcomes.

An engineering result becomes more valuable when another person can inspect it without recreating the author’s private environment. That does not mean copying every local path, credential, and working directory into the report.

The useful handoff preserves identities, observations, scope, and the relationships between them. The private environment should not become the accidental source of authority or the accidental content of a public demonstration.

Name the subject precisely

A path such as “latest capture” is convenient locally and weak as a durable reference. A content identity and a declared generation make it possible to say which bytes a report actually describes. The candidate’s identity matters too: a passing comparison for yesterday’s tree does not automatically apply to today’s edit.

The current KodeBack parity report binds the capture, candidate view, plan, and observation artifacts. It is designed to avoid publishing host origins and machine paths in the report. That is a more useful boundary than asking a reader to trust a screenshot of one developer’s terminal.

Preserve the context needed for interpretation

Portable evidence still needs context. Which profile ran? Which assertions were required? Was the comparison boot-only? Were some observations unsupported? Removing those fields would make the report smaller and less meaningful.

The distinction is between relevant execution conditions and incidental environment details. A declared controlled clock belongs in the interpretation. A private home-directory name usually does not. A report should make the first visible without unnecessarily exposing the second.

Do not confuse portability with universal reproducibility

A report can be inspectable without promising that every reader can run the product on every host. Kalu’s current support remains macOS-specific, and installed-product qualification is separate from the existence of report-producing source code.

Nor should a portable report become a permission slip for mutation. The owner of a later operation must verify current inputs and the evidence appropriate to that operation. A useful artifact can inform a decision without acquiring authority it was never meant to have.

For our public material, this means using synthetic examples and sanitized evidence rather than private donor details. The engineering story should survive that removal. When it does, readers learn the mechanism instead of learning something they were never entitled to know about a project.