Running 15 AI Agents All Day Is Context-Switching Fatigue, IBM Engineer Explains
TejasKumar_ · x · 2026-11-19
IBM AI engineer Tejas Kumar argues that 'AI fatigue' from working with agents is really context-switching fatigue. He runs 10–15 agents in parallel (Claude Ultra + $100 ChatGPT plan); none of the individual tasks is hard — what wears him out is jumping between them all day to answer one, check another, steer a third.
Key points
- His agents handle life admin too: church worship setlists, finding musicians, a Metro card, daily email/calendar digests — many driving browsers via his startup Daiku.
- He cites 2022 peer-reviewed research on task switching and mental fatigue: after 6 hours of hard mental work, people switch to low-effort, sooner-payoff options (scrolling, snacks, skipping the gym).
- Takeaway: supervising an agent fleet drains you through high-frequency switching, not task difficulty — relevant to both agent workflow design and personal energy management.
More from AGI Musings
- Blackstone's Jonathan Gray: Nasdaq at 17x forward earnings, AI is no bubble — sudoraohacker · 2026-11-19
- Ex-SOTA model builder: UK regulation makes pre-training effectively illegal — rosstaylor90 · 2026-11-19
- The AI conscientious objector movement grows on campuses — asusarla · 2026-11-19
- Eric Schmidt: AI race could push countries to bomb each other's data centers — bilawalsidhu · 2026-11-19
- AI may make customer-facing forward-deployed roles more valuable, not less — ivory_tang · 2026-11-19
- Ex-SOTA LLM builder: UK AI is held back by risk-averse capital and 'insane regulation' — rosstaylor90 · 2026-11-19