Overview

AI-generated code optimizes for speed and quantity but can’t address the deeper problem: we still haven’t figured out how to build software well at scale. The gap between “good enough” and “craft” is widening, not narrowing.

The Ellul Critique

Alex Wennerberg’s “AI Code and the Loss of Craft” applies Jacques Ellul’s concept of technique - the reduction of all activity to efficient means toward measured ends. Under technique’s regime, software is “good” if it maximizes metrics with minimal effort. AI agents are the logical extension: they thrive in the “nihilistic space of pure optimization” (Spotify’s algorithmic muzak) and fail where craft matters (Bandcamp’s curated indie scene).

The practical reality matches: AI generates verbose code in a “braindead style” with flat, ugly designs and recognizable aesthetic tells. It works best on well-defined, already-solved problems (unit tests, simple DB functions). Attempts to generalize have “largely failed and produced code that is novel and impressive only in its monstrosity.”

The call to action is a Software Arts & Crafts movement, inspired by Ruskin and Morris’s response to industrialization. When craft becomes more scarce, it becomes more valuable. There’s a treasure trove of unexplored computing paradigms beyond the narrow C/Unix → JavaScript/Web branch we’re on. See Permacomputing.

Engineering Discipline in ML

The Pragmatic Programmer for Machine Learning (Scutari & Malvestio) argues that software engineering practices are critically undervalued in ML. The reproducibility crisis in academia and catastrophic failures in industry (Knight Capital, Zillow’s algorithm) stem from the same root cause: treating engineering discipline as secondary to model performance. Sculley et al.’s insight applies broadly: 90% of the complexity in ML systems is not the model itself.

Connections

• Verification Complexity - Craft and engineering discipline are ways of managing the verification exponent. Sloppy code increases the connectivity factor k.
• AI Economics & the Productivity Paradox - Kent Beck’s programming deflation: the middle disappears, craft and commodity diverge.
• AI Agent Engineering - Ronacher’s lesson that abstractions are premature echoes the craft argument: stay close to the metal until you understand the problem.