What will be left for us to work on?
AI tools are rapidly changing knowledge work, but many argue they are shifting human roles from “doing” to evaluating, judging, and steering rather than eliminating work outright. Commenters share mixed evidence on productivity gains, note that automation historically creates new tasks and demand (often in review, integration, and cleanup), and debate whether software and other white-collar jobs will fragment into tiers of expertise rather than vanish. Underneath this are deeper worries about tech debt, inequality, ownership of AI infrastructure, and whether social and political systems—not technical limits—will determine who benefits if large parts of today’s work become cheap or fully automated.
Shift in Nature of Work
- Many see work moving from “building/doing” to “evaluating, judging, and steering” AI systems.
- Some argue “build wealth before AI obsoletes skills” vs “build skills/agency/judgment” is a false dichotomy; wealth generally requires those skills anyway.
- Others note this division already exists: product/PM/analyst roles vs deeper engineering.
Current Impact of AI on Work
- Several report more work, not less: AI lets them tackle previously “not worth it” tasks (e.g., tech debt, minor bugs, reproductions, small automations). Jevons paradox is cited.
- Some feel work has shifted from coding to requirements, research, and review.
- Others see productivity increases but no reduction in hours; organizations just raise expectations.
- A few concrete job reductions are mentioned (e.g., translation) and smaller teams delivering more.
Tools vs Transformation (Compilers, Low-Code, LLMs)
- Debate over whether AI codegen is “just another compiler” or qualitatively different because of nondeterminism, errors, and the need for judgment.
- Comparisons to previous hype cycles (COBOL, low/no-code, PowerApps): they expanded what developers can do rather than eliminating them.
- Some argue modern agents let you “outsource thinking,” not just implementation, which might be a real break from past tools; others are unconvinced.
Future of Software Roles
- One analogy: software splits into tiers like medicine (few “architect/doctors,” many semi-technical “medics,” and non-technical “nurses” using AI/no-code).
- Critics note these strata already exist (staff engineers, business analysts, power users).
- Some foresee basic coding becoming like literacy: expected across many jobs, with bespoke software work shrinking but high-end roles remaining.
Societal, Economic, and Ethical Concerns
- Worries about inequality and ownership: if a few own the machines, non-owners may have little bargaining power; comparisons to horses replaced by tractors recur.
- Others argue democracies and social policy can redistribute gains; past welfare and labor rights came from political struggle.
- Some fear AI-enabled “Elysium” scenarios or underclass economies; others expect endless “make-work” and new problem domains.
Generational and Cultural Reactions
- Reports that many young people both heavily use AI (e.g., in university) and express hostility toward it; some even “sabotage” corporate AI initiatives.
- Growing “AI fatigue” and suspicion of AI-generated or AI-shaped text; some readers feel much writing now has the same bland AI voice.
Limits, Risks, and Tech Debt
- Several expect a plateau in capability or regulation-imposed slowdown; others think exponential improvement and new architectures will continue.
- Concerns about accumulating AI-generated “slop,” fragile systems no one understands, and future work consisting of cleaning up AI-induced tech debt.
- One large-company case study (internal only) claims all-in engineering productivity gains around ~7%, with some teams seeing negative impact from noisy tools.