Managing up, down, and the robots [audio] (www.swarmia.com)

In a wide-ranging Engineering Unblocked conversation, veteran engineering leader Michael Lopp (Rands) argues that AI is real and useful but currently delivers modest, measurable productivity gains rather than wholesale replacement of human judgment. He praises AI’s “zero-to-one” creativity—how it can rapidly prototype and unblock work—but cautions about hallucinations and the difficulty of reliably measuring value. A key warning: junior engineers who paste unvetted, AI-generated code into projects often slow teams down because they lack the critical thinking and domain knowledge to verify and integrate outputs. Lopp also notes industry-wide noise—companies buying tools without hypotheses—and urges leaders to treat AI like a powerful assistant that still needs human oversight. The episode ties those technical realities back to timeless management practice. Drawing on decades at Netscape, Apple, Slack, Pinterest and Palantir, Lopp is bullish but realistic: leadership fundamentals—empathy, consistent one-on-ones, psychological safety, and listening—remain the biggest drivers of team performance, especially when rebuilding morale after layoffs or scaling past inflection points (he flags cultural shifts around ~150 headcount). He’s even reversed his old “managers shouldn’t code” advice, arguing that hands-on work helps leaders stay credible and spot AI failures. Practical implications: teach juniors fundamentals, require verification of AI outputs, define hypotheses for tooling purchases, and pair AI adoption with strong human processes.