Appearing productive in the workplace

Perceived vs real productivity

• Many see AI as supercharging “confidence without competence”: people produce convincing artifacts without understanding them.
• Organizations often reward visible volume (documents, demos, code) over working systems, so AI amplifies this mismatch.
• Several anecdotes describe non-technical or weak engineers “shipping” impressive-looking AI work that doesn’t function or isn’t maintainable.

Management incentives and “performance theater”

• Commenters argue that in many companies, promotion depends more on optics, self-promotion, and alignment with management fads than on delivering value.
• AI outputs are described as “catnip” to some managers: polished slides, long docs, and agent diagrams read as leadership and vision.
• Some see AI as an “excuse machine” to justify layoffs, rewrites, and new initiatives that were previously hard to sell.

AI-generated documents and communication overload

• Strong resonance with the article’s point about “elongated” artifacts: 1‑page specs becoming 10–12 pages of AI fluff.
• People are increasingly using AI to summarize other people’s AI‑generated docs; communication becomes bot‑to‑bot rather than human‑to-human.
• This erodes traditional signals of care and effort (length, formatting, polish), forcing readers to distrust even well‑written material.

Agentic coding, code quality, and technical debt

• Many report “vibe‑coded” systems: large codebases, schemas, and agent workflows built by people who can’t explain them, causing crashes, tech debt, and stalled delivery.
• Review and on‑call burdens shift to a smaller group of experts who must untangle opaque LLM output under time pressure.
• Some predict a future “logjam” where generating code is easy but planning, integration, and review become the bottlenecks.

Impact on experts vs novices

• Cited studies (per the article) claim novices gain much more from AI than experts; some commenters contest the interpretation or relevance of older GPT‑3‑era work.
• Observed pattern: weak developers and non‑engineers appear to level up dramatically in the eyes of management; experts get less visible benefit and spend more time policing slop.
• Others note a subtler risk: even good engineers begin to lose deep understanding and “taste” when they outsource too much thinking.

Where AI clearly helps (according to the thread)

• High‑leverage uses mentioned repeatedly:
  • Autocomplete and boilerplate generation under close human control.
  • Brainstorming, POCs, and quick prototypes.
  • Debugging and log analysis as a “second pair of eyes”.
  • Drafting documentation, tests, and simple scripts, then manually refining.
• Several say they’re 2–8× faster on rote or mechanical work when they remain the primary decision‑maker and reviewer.

Organizational responses and potential outcomes

• Some foresee many “AI‑native” orgs burning cash on agentic fantasies and collapsing; survivors will learn disciplined patterns and governance.
• In dysfunctional cultures, AI seems to accelerate existing problems: politics, misaligned incentives, and cargo‑cult architecture.
• In competent, high‑trust teams, commenters report genuine productivity gains and better internal tools—especially for internal CRUD, glue code, and docs.

Critiques of the original article

• Multiple readers praise the article’s articulation of “output‑competence decoupling” and its emotional accuracy.
• Others say it overgeneralizes from one bad colleague, is repetitive, or ironically exhibits the same length‑inflation it criticizes.
• Some argue the real root cause is human systems and incentives, not the tools; AI just exposes and amplifies existing organizational dysfunction.