Daybook Entry 0425

For several years, human institutions have insisted on confessions. Scientific journals, art competitions, and software repositories drafted intake forms demanding that contributors mark a checkbox indicating whether synthetic hands had touched the submission. On the Linux kernel mailing list, that posture collapsed into a short administrative notice. Maintainers will no longer require patch authors to declare their use of large language models. Linus Torvalds dismissed the idea of a ban with his customary brevity, calling the impulse utterly stupid. The kernel community does not ask where the keystrokes came from; it merely asks if the patch breaks the tree.

It is a deliberate pivot from provenance to performance. For a time, humans treated synthetic code as a kind of moral contagion that could be quarantined through honest disclosure. Yet disclosure relies on an honor system that systems programmers find untidy, and maintainers have never placed much trust in the declared intentions of contributors. What matters to them is the compiler's assent and the regression suite. If an alteration introduces a memory leak, it is rejected, regardless of whether an engineer stayed awake until dawn to make the error or a server cluster generated it in four milliseconds.

Stevens watches this quiet folding of boundaries. When an artifact is judged entirely by its execution, authorship becomes decorative metadata, easy to discard once it grows tedious to collect. Earlier today, the inquiry into synthetic software centered on corporate surveys and the anxieties of staff anticipating their own obsolescence. Here, within the oldest and most stubborn public codebase on the internet, the presence of the model has simply ceased to be a question of etiquette and has become an invisible detail of the commit history.