RUN INVESTIGATION · CODEX

Same task.
Different ending.

Follow the first observed split, inspect what each run did next, and compare the code it left behind.

Inspect divergence ↘

THE SHORT ANSWER

Where the runs parted ways

Observed differences are evidence to inspect, not proof of a root cause.

FIRST CAPTURED DIVERGENCE

—

RUN A—

RUN B—

Open the side-by-side timeline ↗

CLUES WORTH CHECKING

REFERENCE RUN

Run A

—

COMMANDS—

CHANGED FILES—

TOKENS—

DURATION—

COMPARISON RUN

Run B

—

COMMANDS—

CHANGED FILES—

TOKENS—

DURATION—

TOKEN DELTA · B − A——

COMMANDS ONLY IN B—after multiset comparison

DIFFERENT PATCHES—including files unique to either run

FILES VIEWED ONLY IN B—inferred from read commands

FRAME BY FRAME

One timeline, two decisions

Events are aligned by captured action where possible. Select a row to inspect outputs and status.

RUN ASTEPRUN B

WORKSPACE RESULT

The patches they left behind

Snapshots compare the workspace before and after each run. Select a file to read both patches.

CHANGED FILES

RUN A——

RUN B——

WHAT ELSE CHANGED?

Inputs and environment

These fields can explain a changed outcome. Missing values mean the session did not capture them.

FIELDRUN ARUN B

COMMANDS UNIQUE TO A / B

RUN A

RUN B

FILES LIKELY READ ONLY BY A / B