Snowflake's engineering SVP: ~5% of engineers are AI 'explorers' and you can't predict which 5%

Stack Overflow Blog·medium signalXBlueskyLinkedInCopy link

Vivek Raghunathan, SVP of Engineering at Snowflake, argues on the Stack Overflow blog that engineering orgs split roughly 5% explorers (experimenting past official guidance) and 95% exploiters (shipping with proven methods) — a spectrum, not a binary. The core claim against the 100x-engineer framing: the traits AI amplifies (curiosity, adaptability, willingness to relearn) aren't predictable from prior seniority, so the people posting 100x gains often weren't the most senior engineers before. His recommendation is to let explorers self-identify, extract their discoveries into teachable knowledge, and measure how many engineers move along the spectrum rather than hunting for outliers or hiring them in.