PromptZone 2.0 - Prompts with proof — run before published

Hey Product Hunt 👋 I run PromptZone, and a few months ago I got tired of what every prompt site (including mine) was doing: publishing prompt strings nobody had ever actually run. So I rebuilt it around one rule: nothing gets published on trust. Every prompt is executed in a sandbox against its target model before it's published — the unedited output is shown right on the page — and curated prompts additionally pass review by three independent editorial AI agents from three different model families (accuracy, safety, usefulness). The reviewers reject a lot: while seeding the Marketing & SEO category, they threw out model outputs for omitting a decorative image's empty alt text and for contradicting their own consolidation advice. Those rejections are the product. To be fully transparent: the curation is done by AI agents, not human editors — that's what makes running and reviewing every single prompt possible. What you see labeled is exactly what happened. Live today: careers, coding, and marketing/SEO prompts with evidence on every card, a which-LLM picker rebuilt daily from the live model catalog, and an AI model-release timeline. I'd genuinely value your honest feedback — what's useful, what's missing, which category you'd want tested next. If a prompt fails for you, say so: failed runs are data we act on.