Thesis: Coding agents can reduce the direct problem-solving work that some programmers value, even when they help others build systems more easily.
Observed facts: Willison quotes Scheff describing programming as solving logic puzzles, choosing strategy and timelines, and deciding what was possible. Willison contrasts programmers who mainly enjoy building working systems with those for whom directly solving the problems is a major part of the enjoyment. Mechanism: When an agent performs more of the implementation, it can leave the goal and finished system intact while shifting or removing the puzzle-solving steps that supplied satisfaction. This is a proposed mechanism in commentary, not a measured outcome.
Why now: Coding agents make the division between specifying a result and doing the work toward it newly salient for programming careers. First implication: Teams evaluating agent adoption should consider whether changed task composition affects the work developers find meaningful, not only whether systems get built faster. Evidence boundary: The source is commentary quoting an essay; it provides no survey, employment data, or evidence about how prevalent these reactions are. First watch condition: Look for broader developer accounts or research that tests whether agent-assisted work changes reported job satisfaction or the kinds of tasks people perform.
This starts a distinct analytical record because no prior New Runtime record is supplied; the related earlier discussion points concern coding-agent practice, while this source centers on the subjective value of the work.