Coding agents create professional loss for programmers

Summary

Simon Willison shares and supports Jeremy Scheff's argument that programming agents have eliminated the core intellectual joy of coding—solving difficult logic puzzles for pay—effectively writing a 'death notice' for the craft as many experience it.

Why it matters

As coding agents become mainstream in 2026, this post concretizes a genuine concern among working engineers: the shift from 'I build the solution' to 'I oversee the solution' is changing not just the tools, but professional identity. The post also frames a useful dichotomy—system builders versus puzzle solvers—that helps teams and individuals assess how much of their daily satisfaction is at risk.

Highlights

• Scheff's core claim: Programmers were paid to solve puzzles others couldn't, and large language models (LLMs) have eliminated this scarcity.
• Willison's dual framework: If you enjoy building working systems, LLMs are a net gain; if problem-solving is your source of joy, the loss is real.
• The post is a clear link post with minimal original analysis, functioning more as a signal than a full argument.

Critical take

The 'death notice' framing is hyperbolic for a short link post that essentially amplifies a single author's view. Additionally, the 'puzzle solver vs. system builder' distinction risks oversimplifying the fact that most engineers derive satisfaction from a blend of both. Willison provides no data on how roles are changing, offers no counter-argument, and fails to acknowledge that 'puzzles' may simply be shifting to higher-level architectural and agent orchestration challenges rather than disappearing.