News for hackers. Stuff that matters.
Re:ChatGPT Images 2.0 by vunderba on Friday April 24, 2026 @06:31PM (#47896627)
> it's hard to tell it's wrong without knowing what the prompt was.

Yeah this is actually a huge point of frustration on reddit where lots of people post their "impressive generative images" but fail to disclose the prompts so the audience is only able to evaluate realism/fidelity and not how faithfully the model actually followed the prompt.

There are no comments on this comment yet.