Why effort scales superlinearly with the perceived quality of creative work

Creative iteration: trash vs refine

• Several commenters compare “throw one away” in software to restarting drawings or paintings.
• In practice, both artists and hobby programmers rarely hard‑reset; more often they iterate so heavily it’s equivalent to a restart or abandon and return later.
• Professional artists describe many fast discarded sketches or underpaintings, but almost never scrapping something after many hours—exploration happens early.
• Others highlight studies, thumbnails, and “limbering up” exercises as ways to explore without overcommitting, versus diving straight into a detailed piece.

Skill, cached heuristics, and practice tax

• Multiple comments push back on the article’s suggestion that drawing is always solving a novel problem in real time.
• Experienced artists say they accumulate large libraries of “motor sequences” (hands, poses, constructions) that are recombined and adjusted, analogous to practiced musical licks or coding patterns.
• Some argue last‑mile tweaks don’t help much unless those heuristics were built via thousands of hours of deliberate practice beforehand.

Last‑mile effort, diminishing vs superlinear returns

• Many readers map the thesis to familiar “90/90 rule” or diminishing returns: the last 10% of polish takes disproportionate effort.
• Others note creative work can overshoot: extra polishing can make music overproduced, paintings “turn to shit,” or mixes worse than earlier versions.
• Practical crafts (carpentry, trim, window film) surface a related strategy: design tolerances and overlaps so tiny imperfections don’t matter, instead of endlessly halving errors.

Perceived quality, ranking, and markets

• One line of discussion claims perceived quality is relative and tied to rank in a competitive hierarchy; moving up requires exponential effort because acceptance near the top is narrow.
• Another side rejects pure relativism: people can often tell “much better” from “slightly better” without needing a full comparison set.

Reaction to the article’s rationalist/technical framing

• A large subthread criticizes the abstract as opaque, pretentious “word salad” dressing up a simple idea.
• Some see it as influenced by rationalist/EA jargon that obscures more than it clarifies; others defend modeling creativity with concepts like gradient descent and sample‑space reduction.
• The original poster later concedes the piece was rushed and jargon‑heavy, and clarifies that the intent was to explain why late‑stage edits often fail: the “non‑worsening region” of choices collapses as resolution increases.

Music key / “acceptance volume” example

• The C‑major vs E‑minor example confuses many. Musicians in the thread argue keys are structurally symmetric; the example doesn’t convincingly illustrate different “acceptance volumes,” aside from contextual factors like vocal range or listener familiarity.