Not true at all with frontier models in last ~6 months or so. The frontier models today produce code better than 90% of junior to mid-level human developers.