Nobody Is Thinking Anymore

The piece argues that widespread reliance on large language models has hollowed out real human thinking inside organizations: specifications are half-written by AI, chunks of code are 80% generated, and strategy documents are increasingly AI-derived. That pattern turns LLMs from tools into oracles, letting machine-produced text substitute for the hard, slow work of judgment - sitting with problems, arguing trade-offs, listening to critique, and making imperfect but accountable decisions. Treating summaries or drafts as equivalent to understanding creates a false sense of competence and risks delegating choice about what to build and prioritize to statistical predictors rather than people who must live with consequences.

The practical consequences are skill atrophy and loss of the drills that cultivate taste and judgment. Automating routine steps removes the learning loops that teach engineers why systems behave the way they do. The remedy proposed is procedural: insist newcomers produce their own explanations, keep friction in the process so thinking can incubate, and use AI only as a copilot while people retain responsibility for deciding what to build. The central claim is clear and actionable: preserve human deliberation and accountability before outsourcing the very judgments that define direction and value.