Crayfish Carnival, "Coding Freedom": Why Do We Feel More Exhausted as AI Gets More Powerful?

Faster outputs, flatter energy

Developers and knowledge workers are reporting a paradox: more powerful AI is speeding up tasks but leaving people more exhausted. It has been reported that former Google and Amazon engineer Steve Yegge described a “dozing attack” after long sessions of vibe coding, and Siddhant Karel reportedly wrote that tasks that once took three hours now take 45 minutes — yet he feels more tired than ever. The hallmark is not slower work; it is an accelerated tempo combined with fragmented attention and rising invisible labor.

Evidence from the field

It has been reported that researchers at the University of California, Berkeley’s Haas School of Business spent eight months embedded in a roughly 200‑person U.S. tech firm and found the same pattern: AI tools quickened workflows but extended work hours, increased cognitive fatigue, and worsened decision‑making after the initial novelty wore off. Business press and academic commentary — from Harvard Business Review to Business Insider and Forbes — echo these findings, noting that prompt‑tuning, continual debugging, and 24/7 parallelized outputs push people into constant task switching rather than deep focus.

Why it happens, and what to do

Why does productivity feel like exhaustion? Because using AI often creates hidden work: learning new models, integrating tools, refining prompts, and reviewing others’ AI‑generated output. When the barrier to “can we do it?” falls, the question becomes “should you do it?” — and social or managerial expectations quickly expand job scopes. The practical fixes are already being suggested: use AI with a task‑first mindset, set time limits on tuning, accept “80% good enough,” schedule non‑AI deep‑work days, and have candid conversations with managers about scope creep. If unregulated, AI won’t just burn CPU cycles — it will burn human attention.