If you were dropped into a coding environment where the underlying model was hidden (GPT-x, Claude, Gemini, grok, etc) do you think you could reliably tell which one you were using or at least which family?
In other words... after all this vibe coding could you identify the model strictly off vibes?
If yes, how long would it take you to be confident? And what constraints would you need for the test to be meaningful (i.e. familiar codebase vs greenfield, real bugs vs toy tasks, time-boxed, language/framework, etc)?
0 comments
—
Ask HN: In a blind coding test, could you identify an LLM strictly off vibes? · HackerTrans
In other words... after all this vibe coding could you identify the model strictly off vibes?
If yes, how long would it take you to be confident? And what constraints would you need for the test to be meaningful (i.e. familiar codebase vs greenfield, real bugs vs toy tasks, time-boxed, language/framework, etc)?