In a recent essay, Eugene Yan offers a metaphor to guide one of the most common career questions in tech: should I join a start-up, and which offer should I accept? He casts the options as commando, soldier, and police. Commandos operate in small teams with unclear missions; soldiers execute clear, large-scale plans; police enforce order and minimise risk. Each role rewards a different kind of person and a different phase of career.
For machine learning professionals, the metaphor is a useful way to think about team environments. A commando-like role might mean open-ended research and rapid prototyping in an early-stage company. A soldier role could mean building and maintaining ML systems inside a larger organisation with established processes. Police work might appeal to those drawn to safety, reliability, and responsibility for preventing costly errors.
The value of the essay is not in declaring one path superior. Instead, it frames the decision around personal fit and conditions for growth, encouraging readers to ask what type of environment will help them develop at a given moment. That is a more useful question than simply comparing compensation or company prestige.