However, Grok outperforms on logic tasks—software being logic at scale—when those patterns aren't covered by Opus's pretraining
For example, the Tesla chip design team prefers Grok to Opus even after being asked to try Opus