I have 5 models the same open-ended prompt every day for a month. 6000 people played the results ...
retained text
The setup: Claude, ChatGPT, Gemini, Grok and Muse each get the same one-paragraph prompt once a day (scheduled task). Build one self-contained HTML app, no libraries, your choice of what to make, and it can’t repeat anything you’ve built before. No follow-ups or fixes. Whatever comes back gets published as is (With a few exceptions which I had them iterate a few version of) That’s 104 apps so far. Over the last four days about 6,000 people opened them roughly 20,000 times and left 147 likes. Likes per app, by model: * Claude: 2.6 (31 apps) * Muse: 1.3 (27 apps) * ChatGPT: 1.1 (19 apps) * Grok: 0.8 (12 apps) * Gemini: 0.2 (15 apps) A couple of caveats: \- I did have them focus on productivity and business apps for a week but I then removed that restriction as it was getting very boring. \- Everything is playable in the browser, grouped by day so you can compare what each model did with the same prompt. The full prompt is on the page, and there’s a “Surprise me” button if you just want to poke around. \- A couple of these were not part of the scheduled jobs and they were prompted specifically to do what they're doing (AWS Diagram Studio ,Label sheets ) \- Also, Sonogram, Kaleido and Ripple had the most post prompting of any of them. Most of the others were published as generated. Not every day has an app generated due to insufficient credits, or errors, or whatever. You'll get the idea.
document fetched 2026-11-18 00:28 · canonical https://reddit.com/r/OpenAI/comments/1xyzkoo/i_have_5_models_the_same_openended_prompt_every