ML PhD Student's 10-Hour Schedule and Coding Agent Pressure
A top-5 ML PhD student reveals anxiety without Slurm jobs and agent pitfalls.
A viral Reddit post from a 3rd-year PhD student at a top-5 US ML program offers a raw look at daily life in machine learning research. They work 9–10 hours a day in broken chunks: a focused morning block, afternoon meetings, evenings for commute/exercise/socializing, and additional deep work late at night. Weekends mix errands with at least a little work daily. A key anxiety driver is Slurm job scheduling — the student feels uneasy without at least some jobs running to collect results later.
Beyond scheduling, the student highlights growing pressure to use coding agents (AI coding assistants) but finds them double-edged. Agents create 'dead time' where the student waits for them to finish thinking, fostering an illusion of productivity while actually reducing active engagement. This trade-off reflects a broader tension in AI research between leveraging automation and maintaining deep focus. The student targets top ML conferences (NeurIPS, ICML, ICLR) and core NLP venues, and struggles with consistently making deadlines.
- 9–10 hour workday is non-contiguous, with focus peaks in morning and evening.
- Feels anxious without Slurm jobs queued; schedules them to run offline.
- Coding agents create dead time and illusion of productivity, reducing active work.
Why It Matters
Highlights real productivity trade-offs and mental health challenges in ML research, relevant to remote work and AI tool adoption.