Train a decent model, but market with wrong emphasis. Discussion derailed by people doing napkin maths of how it doesn't make sense than actual model discussion. Classic marketing mistake, classic HN response.