AI-generated “workslop” is destroying productivity?
Limits of AI Understanding and “Tapeworm” Content
- One subthread argues LLMs can’t grasp high-dimensional, event-based meaning (memes, paradox, rich cultural references), only low-bandwidth token patterns.
- “Tapeworm format” is described as non-causal, contradictory events with potentially infinite interpretations (Koans, complex art, meme chains) that resist compression into simple semantics, and thus resist automation.
- Others push back that humans also often don’t know what things “really mean,” so the bar being set for AI is unrealistically high and the critique drifts into jargon.
AI Code Slop and the Cost of Review
- Multiple stories of non-technical managers or juniors pasting large AI-generated pull requests: huge, convoluted code for simple CRUD tasks, cache hacks, etc.
- Reviewers report that refuting or cleaning this is far more work than writing the feature properly, invoking Brandolini’s law.
- A recurring point: reviewing AI code is harder because there is no underlying intent to recover; you must suspect every line.
- Some engineers report a productive pattern: use AI for a rough “vibe” solution, then rewrite it cleanly by hand using that as a sketch.
Corporate Mandates and AI Hype
- Many describe management mandating AI use and even mandating that it “make you more productive,” with performance reviews requiring examples of gains.
- Critics see this as pre-ordaining the answer and manufacturing justification for sunk AI spend, akin to Stakhanovite/metrics theater.
- Some managers admit they see no real cost savings or margin improvement despite heavy AI use, especially in maintenance/extension work, but hype and C‑suite pressure persist.
Workslop in Docs, Meetings, and Communication
- AI-generated emails, reports, PRD prototypes, and meeting notes are described as polished but substantively wrong, verbose, or incomplete.
- New pattern: bullets → AI-fluffed prose → AI-summarized back into bullets; “slop human centipede.”
- People report executives and managers thrilled with long AI reports that are factually weak, shifting verification burden downstream.
Bullshit Work, Arms Races, and Nominal vs Real Productivity
- Several link “workslop” to existing bullshit work: decks, reports, and notes no one really needs. AI just lets people produce more of it, faster.
- Fear of an arms race: AI to generate junk, AI to parse junk, AI to summarize the parse, burning energy while adding little value.
- Some frame this as nominal productivity (more artifacts) rising while real productivity (useful outcomes) stagnates or falls.
Authenticity, De-skilling, and Personal Use
- Concerns about people outsourcing thinking and losing skills (navigation via GPS, writing via LLMs).
- Tension around personal writing: AI-assisted memoirs may help someone express themselves, but readers may feel the author’s “voice” is lost.
- Several note that AI is good at generic filler; the hard part—the original thought, judgment, and responsibility—remains human.