Strategic insights on technology leadership, engineering culture, and building great teams
Agent systems can behave in ways nobody explicitly chose. Making their decisions visible helps, but we also need to observe how the parts interact. Biology offers some useful patterns for keeping those systems reliable.
AI gives us the chance to redesign how we build products around outcomes. Applying it to every stage of the existing software development lifecycle would preserve too many assumptions from a world where implementation was expensive.
AI hasn't removed the need for managers' judgement and attention. I think we're taking away the time for both by expecting managers to code more while keeping the rest of the job.