Editing as the Core Creative Act
The received image of the author is a person generating text, words flowing from mind to page. In this image, editing is secondary — a cleaning-up operation performed after the real work is done. Becker knew this image was wrong because the people who actually make things describe writing as a process of selection, not generation from nothing. Jazz musicians describe improvisation the same way: real-time editorial choice among phrases that present themselves, rejecting most, shaping the few that survive. Photographers describe their work as editing with special emphasis: hundreds of exposures, handful selected. The ratio of rejected to accepted is enormous. The editing is the art. This observation has radical implications for AI-assisted creation: if the core creative act is selection rather than generation, then a tool that generates at unprecedented volume does not displace the creator — it intensifies demand for the creator's actual skill, which was never generation in the first place.
In the AI Story
When producing text, code, images, or music costs almost nothing in time or effort, the constraint shifts from production to evaluation. The scarce resource is no longer the ability to generate material but the ability to judge whether the material is any good. This shift has precedents in every art world that has undergone a production revolution — recording technology, desktop publishing, digital photography. Each time, the creative premium relocated from generation to editing.
Segal describes his discipline with Claude precisely: 'the willingness to reject Claude's output when it sounds better than it thinks.' The formulation is exact. The output sounds right — grammatical, logical, reference-complete. But sounding right and being right are different things, detectable only by someone who knows enough about the subject to distinguish plausible from true. The Deleuze episode in The Orange Pill — where Claude drew a connection between flow state and a concept misattributed to Deleuze, elegantly and wrongly — is the paradigmatic case.
The AI world's editorial conventions are under-developed relative to the volume of production the tools enable. A builder working alone has no code reviewer, no peer reviewer, no director watching from the house seats. The editorial function must be performed by the person who generated the material, creating a conflict of interest that every mature art world manages through structural separation: publishing houses employ editors who are not authors, orchestras employ conductors who are not composers.
Several editorial conventions are competing in the AI world. Iterate-until-correct, applying domain knowledge at every step. Generate-review-ship, maximizing speed and volume. Structural separation, replicating the mature art world's division of labor. Which convention stabilizes will determine whether the AI world produces work that is merely abundant or work that is actually good. The aesthetics of smoothness will become the default if unedited generation is the norm.
Origin
Becker developed this observation across his studies of photography (where he worked as a photographer and theorist), jazz, and writing. He formalized it most systematically in Writing for Social Scientists (1986), which argued that the first draft is raw material for the real work of composition: cutting, reshaping, discovering what the piece is actually about by eliminating what it is not.
The photography work was particularly formative. Becker edited his own photographs and studied how other photographers edited theirs, observing that the contact sheet — with its hundreds of rejected exposures — revealed the work of selection that the single exhibited print concealed.
Key Ideas
Generation is the raw material; selection is the art. Every creative practitioner, when observed carefully, turns out to be editing more than generating.
Production revolutions shift the constraint from generation to evaluation. Recording, desktop publishing, digital photography, and now AI have all relocated the creative premium from producing material to judging it.
Self-editing is structurally difficult. Mature art worlds evolved separation-of-function conventions because the person who generated material has attachments the external editor does not.
Smooth output is the new mediocre. When competent production is universal, competent-looking becomes indistinguishable from competent-looking, and the bar shifts.
Editorial conventions determine what reaches an audience. Strong editorial conventions sustain quality; weak ones allow the plausible-but-hollow to proliferate.
Debates & Critiques
Some argue that AI tools will eventually perform their own editing through recursive self-refinement, rendering human editorial judgment unnecessary. Becker's framework suggests this confuses two kinds of editing: the technical refinement of output against specified criteria (which AI can approach) and the judgment about what the criteria should be in the first place (which requires stakes in the outcome that AI does not have).
Appears in the Orange Pill Cycle
Further reading
- Howard Becker, Writing for Social Scientists (University of Chicago Press, 1986)
- Howard Becker, Do You Know...? The Jazz Repertoire in Action, with Robert Faulkner (University of Chicago Press, 2009)
- Donald Murray, 'The Maker's Eye: Revising Your Own Manuscripts' (The Writer, 1973)
- John McPhee, Draft No. 4: On the Writing Process (Farrar, Straus and Giroux, 2017)
- William Zinsser, On Writing Well (Harper & Row, 1976)