Developers report burnout from LLM-driven workflows

Developers are describing a new kind of work pressure created by LLMs: a constant stream of generated code, documents and tasks that still require human judgment. Several commenters said AI has made it easier to start more work than they can responsibly finish, leaving them to manage queues of agent output, parallel projects and follow-up reviews rather than spending long stretches on focused engineering.

The review burden emerged as a central complaint. One developer described being sent about 30 LLM generated documents for immediate review, while another said AI-generated slides that took a boss 5 minutes to create could require 3 hours of fact-checking and rewriting. Others said teams are producing low-quality pull requests, business documents and plans faster than colleagues can verify them, shifting labor from creation to validation.

Workplace incentives are intensifying the problem in some organizations. Commenters cited token leaderboards, time-spent measurements and performance review criteria tied to “effective use of AI,” creating pressure to show visible AI usage even when it may not improve output. Some compared the shift to automation that removes the creative part of a craft and leaves repetitive QA or rubber-stamping.

Not everyone saw the change negatively. Some developers said agents helped them ship more ambitious projects, improve tests and clear old backlogs, with one solo developer saying output had gone up 20x. Suggested mitigations included requiring people to review their own AI output before sharing it, limiting concurrent agent sessions, using style guides, strengthening tests and treating verification as a first-class part of the workflow.