Drawing on insights from Snowflake's SVP of Engineering, this piece challenges the common leadership instinct to identify and clone '100x engineers' who excel with AI coding tools. Using a reinforcement-learning-inspired explorer/exploiter framework, it argues that roughly 5% of engineers are natural explorers who push AI tools to their limits, while 95% are exploiters who prefer paved paths. The real mistake is treating this as a binary rather than a continuum. Effective AI adoption strategy means letting explorers self-identify, building structured mechanisms to move the middle of the org up the scale, measuring that movement rather than counting outliers, and giving exploiters credit for prioritizing delivery over experimentation.
Table of contents
A continuum, not a cast of charactersWhy the obvious moves don’t workWhat actually needs managing hereThe question worth sitting with97 Impressions