- Software engineer Siddhant Khare gained attention with his essay “AI fatigue is real and nobody talks about it,” warning about “AI fatigue” in the tech industry.
- Khare stated that AI tools help him write and deploy more code than at any point in his career, but also leave him unprecedentedly exhausted.
- An engineer’s job shifts from “builder” to “reviewer,” constantly screening pull requests and AI-generated results on a never-stopping assembly line.
- AI reduces the cost of code production but increases the cost of coordination, evaluation, and decision-making, a burden placed almost entirely on humans.
- A typical workday for Khare involves context switching between about 6 different problems, each taking “only an hour with AI,” but collectively causing cognitive overload.
- He describes: AI doesn’t get tired between tasks, but humans do.
- Many other engineers on X, Hacker News, and Lobsters empathized, describing burnout from “vibe coding” and waiting for agent responses then correcting them.
- An 8-week Harvard Business Review study with 200 tech employees showed that AI did not reduce work but increased workloads, leading to cognitive fatigue and declined decision quality.
- Khare also faces FOMO pressure to constantly track updates from OpenAI, Anthropic, and other AI firms, even using weekends to try new tools.
- He worries about skill degradation, as the ability to reason and write code manually weakens with over-reliance on AI.
- Andrej Karpathy also admitted he is starting to lose his ability to hand-code.
- To save himself, Khare set rules limiting AI usage and took a temporary “AI detox” during a 14-day vacation.
- He suggests AI companies should design more “guardrails” so users don’t push themselves to exhaustion.
📌 The lesser-known downside of AI: it boosts productivity to record levels but trades it for mental fatigue, burnout, and skill degradation. In the software industry, AI reduces code production costs but increases coordination, evaluation, and decision-making costs, burdens that fall almost entirely on humans. Engineers shift from “builders” to “reviewers,” constantly screening AI-generated results on an endless assembly line. Humans face pressure to constantly track updates from OpenAI, Anthropic, and other AI firms, even using weekends to test new tools.
