We mourn our craft

Diverging attitudes toward AI-assisted coding

• Thread splits between those thrilled by “agentic engineering” and those grieving the loss of hands-on coding.
• Enthusiasts say LLMs remove drudgery, accelerate learning, and let individuals build things previously out of reach.
• Skeptics feel reduced to “LLM PR auditors” or “glorified TSA agents,” finding prompting and code review less satisfying than writing code.

Craft, joy, and identity in programming

• Many describe coding as a craft akin to woodworking, music, or painting: pleasure in repetition, small design decisions, and “holding code in your hands.”
• Others say their real joy is making useful things; code was always just a medium. For them, tools changing is fine as long as building remains.
• Some fear loss of community and shared “war stories” as fewer people deeply engage with low-level details.

Productivity gains vs quality and “slop”

• Supporters claim LLMs handle boilerplate, shell scripts, scaffolding, config, test generation, and mundane data plumbing with big productivity wins.
• Critics highlight hallucinations, brittle code, unreadable patterns, duplicated logic, and increased outages/CVEs; they see a “slopification” of software.
• Concern that non-experts will ship “looks like it works” systems with hidden security and scaling failures, while maintainers bear the cost.

Natural vs formal language; determinism and trust

• Several stress that we invented formal languages precisely because natural language is ambiguous; “natural language programming” is seen as inherently imprecise.
• Compilers are deterministic and well-understood; LLMs are probabilistic black boxes whose behavior is hard to reason about or fully verify.
• Some push back that real-world software is already messy and non-perfect, and rigorous testing is needed either way.

Careers, juniors, and labor market anxiety

• Strong worry from younger devs and students: they just entered the field as LLMs arrived; they fear devalued skills and shrinking opportunities, especially for juniors.
• Older devs with savings tend to be more relaxed, sometimes exiting or shifting roles; others feel the timing robbed them of a once-aligned passion and career.
• Debate over whether juniors become more valuable (augmented learners) or redundant (LLMs replacing entry-level work).

Power, centralization, and social impacts

• Many object less to the tech than to its control: a few corporations owning critical models, data, and hardware; dependence on subscription “thinking as a service.”
• Fears of broad white‑collar job erosion, worsening inequality, and a “techno‑feudalist” future where labor has little bargaining power.
• Some see historical continuity with past automation (Luddites, industrialization); others argue this time is different because cognition and creativity are being targeted.

Historical analogies and “six months” skepticism

• Repeated comparisons to photography vs painting, synthesizers vs musicians, woodworking vs CNC, self-driving cars, and past overhyped tech.
• The mantra “wait six months” is heavily criticized; people note moving goalposts and lack of visible, robust, high‑quality AI‑built systems at scale.

How LLMs are actually used today

• Common positive uses: shell scripting help, translation between languages, refactors, infrastructure boilerplate, test scaffolding, debugging large logs, quick prototypes.
• Many describe a hybrid workflow: humans design architecture and key logic, use LLMs for drafts, then heavily edit and review.
• There’s broad agreement that LLMs are far from reliably doing full-stack, production-grade systems without strong human oversight—though opinions diverge on how fast that could change.