In an overheard anecdote, AI commentators react with amusement to a claim that early OpenAI Codex authored only 20% of code

No technical details or benchmarks supported the claim