After AI takes everything

AI-generated code quality and “slop”

• Many commenters report that LLM-written code is often subtly wrong or structurally poor, even on simple tasks (e.g., a byte-based LRU implemented as entry-count-based).
• Review cost frequently outweighs generation speed; experts say it’s often faster and safer to write the code themselves.
• Others reply that human code is also buggy and duplicated; they argue LLM output only needs to reach “good enough,” not elegant.
• Concern that managers can’t recognize slop, but will still drive adoption, accelerating software “enshittification.”

Taste, craftsmanship, and responsibility

• Several people frame the difference as “having taste” or “giving a shit”: clear mental models, consistent structure, and willingness to clean up.
• LLMs in unskilled hands magnify bad taste; some see this as a deserved existential crisis for shallow practitioners.
• Others counter that elegance has always taken a back seat to shipping; AI just makes this more visible.

Timelines and hype skepticism

• Strong pushback on claims that AI will “take everything” in 18–24 months; predictions are seen as unfalsifiable hype.
• Some argue model improvements are now incremental, not paradigm shifts.

Jobs, skills, and economic impact

• Debate over whether most work disappears vs. shifts to new roles (skilled trades, healthcare, deeper-stack engineering).
• Worry that even “AI-proof” jobs will be flooded by displaced white‑collar workers, depressing wages.
• Core anxiety: not just losing a job, but losing perceived worth and the ability to support a family.

Who benefits and how to respond

• Repeated question: if AI generates massive wealth, who captures it—few elites, or society broadly?
• Some see parallels to past tech revolutions; best case is intense competition driving AI costs down and benefits to consumers.
• Others stress organizing and policy (copyright reform, mandatory licensing, redistribution) over individual “hustle” narratives.

Data, ownership, and paying humans

• Idea that AI is parasitic on human-created data; suggestions that models should pay individuals for content usage or fund large pools of creators.
• Counterpoint that trying to assign per-person value becomes equivalent to taxation/redistribution problems.

Meta: AI-written essays and discourse fatigue

• Several think the linked essay (or its English version) shows LLM hallmarks: bloated length, repetitive slogans, over-styling.
• Frustration with long, philosophical “how to adapt” pieces that seem to skip over concrete, material questions.