In his recent essay, machine learning engineer Eugene Yan shares a personal productivity system built around a lean stack of tools and deliberate routines. He emphasizes that the goal is not to add more apps or hacks, but to reduce friction so that the bulk of energy goes into actual creation—writing, coding, and analysis. For AI/ML practitioners, this means treating your own workflow as a system to be optimized, much like a model pipeline.
Central to his approach is the concept of "autopilot routines"—fixed, low-effort habits that handle recurring tasks like email triage, note capture, and reading lists. By codifying these decisions, he frees up cognitive load for the more demanding parts of research and engineering. He also advocates for the "Maker's schedule," which carves out long, uninterrupted blocks of time for deep work, as opposed to a manager's schedule of back-to-back meetings.
What's refreshing is that his suggestions are grounded in personal experience, not generic productivity advice. He acknowledges trade-offs and doesn't pretend a single tool works for everyone. The underlying principle—accomplish more with less by designing intentional defaults—is directly applicable to machine learning workflows, where iteration speed and focused attention often matter more than raw hours.