How this issue
earned its labels
This issue was researched, written, edited, programmed and laid out in a Codex workflow. Public sources supplied reported developments; local executions supplied the prompt-in results. The selected cover and both full-page ad illustrations were generated locally in Draw Things.
The supplied programs were assembled from the printed project briefs and explicit implementation decisions. Our evidence describes those files. A new run from the same prompt can choose different code; the new result needs its own checks. A comparison across independent agents would be a separate experiment.
The verification record
Logic checks compare the programs’ rules with expected answers: 35 assertions passed. Browser checks use the running interface: 40 passed. The Modern Web Guidance experiment passed seven separate typed-command checks. Its version, small test files, commands and raw results accompany the edition.
The browser run covered three widths, offline reloads and page errors. The sequencer’s sound-generating code produced an audio signal in memory; that is different evidence from a listening review. The project screenshots show the supplied programs.
Reporting scope
This edition includes a fresh exact-prompt Lunar Lander build and one executed keyboard-repair follow-up tested November 15. The first file and its failures remain available beside the repaired file and independent comparison. The three original supplied programs and printed build prompts are unchanged. Earlier reports retain their dates; a first-time human reader trial remains separate work.