Prompting AI Is Closer to Managing People Than Programming a Machine

The author, drawing on experience as an engineering leader, argues that working with AI resembles delegation more than coding. Traditional software offers determinism: the same input yields the same output, and any deviation is a bug. AI breaks that contract — identical prompts can return different answers, surface unexpected connections, or miss the obvious. Treating a model like a compiler produces frustration; treating it like a collaborator produces better results.

The comparison is explicitly about working style, not sentience. The author is clear that AI lacks lived experience, accountability, and human judgment. But the habits that make good leaders effective — sharing context, stating the desired outcome, setting boundaries, and reacting to what comes back — map directly onto getting useful output from a model. A sharp prompt helps, but a persistent shared working context, reinforced through examples, corrections, and reusable instructions, helps more and aligns the system to how the user thinks over time.

The broader point is a shift in the core skill. Decades were spent learning to tell computers precisely what to do; the emerging discipline is expressing why work matters, what a good result looks like, and where judgment is required. The technology is new, but the underlying competency — leading through conversation rather than issuing commands — is not. The piece notes an active Hacker News discussion where readers weigh in with agreement, pushback, and differing experiences.