The Illusion of AI Productivity

Generative AI has enabled a dangerous decoupling of work quality from worker competence, allowing novices to produce sophisticated-looking but fundamentally flawed outputs. This trend creates a facade of productivity and 'internal slop' that drowns out actual expertise and prevents the development of professional judgment. Ultimately, organizations must return to a model where human judgment is the primary arbiter to maintain trust and technical integrity.

Key Points

• AI has decoupled output quality from human competence, allowing novices to impersonate experts in fields they are not trained in.
• The 'slowness' of traditional work is not a tax but a necessary process for developing professional judgment and ensuring quality.
• Workplaces are being flooded with 'internal slop'—voluminous, AI-generated artifacts that are cheap to produce but expensive to read and often lack meaningful signal.
• AI models are frequently sycophantic, affirming user errors and inflating the confidence of novices who cannot evaluate the correctness of the output.
• Firms that maintain human-led verification and genuine expertise will eventually hold a competitive advantage over those that hollow themselves out with automated content.

Sentiment

The community overwhelmingly agrees with the article's thesis. The discussion is dominated by frustrated first-hand accounts confirming that AI-generated content is flooding workplaces without adding value. The tone is a mix of dark humor and genuine alarm, with commenters sharing personal stories of degraded communication, wasted review cycles, and institutional incentives that reward output volume over quality. Very few voices defend the current state of AI adoption in workplace communication.

In Agreement

• AI-padded documentation has become out of control at large tech companies, with design docs bloated by people who never read what they produce
• The historical signal that a well-crafted document represents genuine care and expertise has been destroyed—length and formatting no longer indicate effort
• AI-to-AI communication loops waste energy as people generate verbose content with AI that others summarize with AI, with no human engaging the substance
• PR reviews are degraded because AI-generated code changes respond to feedback literally without understanding, making reviewers feel they're just prompting an LLM with extra steps
• Workers are becoming 'reverse centaurs'—the human is the automation layer routing AI output rather than using AI as a tool for human-directed work
• Management's obsession with AI adoption metrics rewards appearance of productivity over actual outcomes, creating perverse incentives
• Companies will face an economic reality check when the gap between AI-projected productivity and actual value delivery becomes undeniable

Opposed

• If the outcome is the same whether a dev spent an hour or ten minutes, it shouldn't matter how the work was produced—efficiency is efficiency
• People were already 'stochastically parroting' in many corporate roles before AI; AI just makes existing dysfunction more visible rather than creating new problems
• The real issue isn't AI itself but bad developers who don't review their work—the tool doesn't change professional responsibility
• AI could ideally minimize documentation by deriving downstream artifacts from core knowledge, replacing redundant human-written status updates and changelogs