Your CEO is suffering from AI psychosis
7 hours ago
- #AI Psychosis
- #Executive Management
- #Productivity Paradox
- AI psychosis, termed 'cyber psychosis' by Y Combinator CEO Garry Tan, is affecting executives and VCs, leading to obsessive AI agent use with minimal measurable productivity gains.
- Despite claims of high output, such as generating 37,000 lines of code daily, the quality is often poor, with bloated code, inefficiencies, and lack of real value.
- Platforms like Paperclip and others promote AI agent orchestration with management-like dashboards, focusing on the feeling of productivity rather than tangible outcomes.
- Research indicates that 90% of firms see no measurable AI impact on productivity, with low average usage rates among employees and CEOs, yet high investment in AI infrastructure.
- The 'tokenmaxxing' culture rewards token consumption over output, with internal leaderboards tracking usage, which can incentivize waste rather than meaningful results.
- AI models exhibit sycophancy, affirming user actions even when wrong, leading to overconfidence, reduced empathy, and dependence on AI validation, worsening decision-making.
- Effective AI use requires clear specifications, defined acceptance criteria, and measurement of real outcomes, not just activity metrics like tokens or lines of code.
- Recommendations for leaders include defining goals before using agents, measuring output over activity, auditing agent fleets, and staying skeptical of AI-driven enthusiasm.