1. Feed different LLMs the help center article (all the examples) and feed them the example from the JAI preset script. 2. Be specific with them that JAI wasn't SillyTavern and fed anything, even the small "instruction" shown on the JAI page. 3. Check the JAI script page context and record down all things and feed AI again. All these steps look exhausting, I know, but it's the only way to make LLMs not mistake your final goals. And you still need to pay attention in between. Then compare the answer they give and ask more questions. Don't directly ask AI to write for you the script without the basic info I listed out just now, because the JAI script is not working; that's smart for now. So feeding LLMs all the information you have is to prevent them from making up things too (and yourself from learning when reading). Ask around in the JAI server in the script channel, but read the whole thing they put out there first. Ummm, I know this sounds intimidating to execute? But that's how you learn. And the reason I tend to use different models on the same questions is because I'm not trusting everything, and I'm a control freak. XD The models I use are Grok, Gemini, DeepSeek, Claude, Copilot (just sometimes), GPT, and Perplexity.* I have zero background or experience in coding, and I hate to always bother people before trying it myself; that's why I got all this tangled up at once. I'm not sure I have given you the answer you want or not XD