Toward a Software Arts & Crafts Movement

AI flourishes in metric-obsessed systems, generating “good enough” output that crowds out craft—just as Spotify’s ecosystem favors algorithmic blandness while Bandcamp rejects AI. Big-tech software mirrors this problem: narrow roles and bloated processes produce low-quality code, and while AI can automate rote tasks, it fails at deep understanding and open-ended design. The author calls for a software Arts & Crafts movement to restore human-scale craft and explore alternative computing paths, whose value will grow as AI slop spreads.

Key Points

• AI thrives in environments governed by technique—where success is measured by engagement and efficiency rather than craft—producing abundant but shallow content.
• The music platform contrast (Bandcamp vs. Spotify) illustrates how metrics-driven ecosystems incentivize algorithmic blandness and invite AI, whereas craft-centered spaces resist it.
• Modern software engineering in large companies has lost craft, becoming narrow, bloated, and low-quality; AI fits this regime by automating rote tasks but cannot address the deeper quality crisis.
• AI agents are useful for well-defined, repetitive problems but routinely fail at open-ended design, deep understanding, and non-mainstream computing contexts.
• A software Arts & Crafts movement—reviving older ideas, human-scale design, and alternative computing paths—can restore craft, whose value will rise as mass-produced software proliferates.

Sentiment

The community is moderately sympathetic to the article's thesis. The dominant framing — that LLMs amplify existing tendencies for better and worse — broadly aligns with the article's concern about craft degradation. However, a substantial pragmatist camp pushes back, arguing AI is simply another tool and that code quality was already declining independent of AI. The discussion is thoughtful and substantive rather than dismissive, with the article's author participating to refine points. The strongest agreement clusters around the efficiency-as-ideology critique, the self-delusion problem in AI-assisted coding, and the junior pipeline concern.

In Agreement

• Efficiency has been tacitly elevated to society's highest unchallenged value, and Goodhart's Law applies — what gets measured (engagement, velocity) becomes the goal, crowding out craft and care
• Code quality was already declining before AI — business treats software as a commodity where bugs are an accepted cost, and AI simply fills the niche where poor practices already existed
• LLMs are an amplifier that makes lazy developers produce more slop, and even experienced developers suffer self-delusion about maintaining quality when using AI
• The junior developer pipeline is at risk — if AI lets seniors do the work of multiple juniors, companies will drop entry-level roles, but then no one develops into the next generation of seniors
• Developers have influence rather than control over LLM output, and overselling AI capabilities through flawed team-management analogies will harm long-term adoption
• AI-generated Forth code demonstrates the article's point about forgotten computing paths — LLMs produce code that reads like translated C rather than idiomatic Forth
• The Luddite parallel is apt but deeper than commonly understood — the original Luddites were a labor movement, and this disruption may be qualitatively different because AI is a general cognitive labor substitute

Opposed

• High-quality code can be produced through AI agents via disciplined orchestration of planning, implementing, validating, and reviewing — the craft shifts to directing the tool rather than typing the code
• AI is simply another tool like CNC machines, power washers, or chainsaws — the craft lies in knowing when and how to use it, and historical analogues show craft persists alongside automation
• If AI makes bad software cheaper, that is fine — what matters is whether we also get more great software in absolute terms, which seems likely as developers are freed for higher-quality work
• A puzzle craftsman and guitar luthier both use CNC and AI tools to expand creative capability — refusing to use available tools is not craftsmanship but stubbornness
• Popular software is average by definition because it must appeal broadly, so lamenting lack of craft in industrial software is pointless — AI is just another tool in that industrial process
• The article's logic is incoherent for conflating choice of tools with product quality decisions — tool choice does not determine whether the result has craft