OpenAI Codex engineering lead Thibault Sottiaux sparks debate on whether practitioners trust standard benchmarks over peer recommendations · Digg