Unlike technical debt, which shows up as harder-to-maintain code, cognitive debt shows up in the developer's head. Ankur Sethi's August 2026 proposal to manually retype every LLM-generated edit sparked a Hacker News debate over whether retyping builds comprehension or just slows acceptance. Critics argued active review, design-first delegation, and testing matter more than mechanical transcription; the underlying risk tracks how passively a developer accepts output.