My experiment with an ultra-compact bot focused not on providing factual knowledge to the generation engine, but rather on giving it context and structure in a general way about the logic of the universe in which it will operate. I discourage using JLLM; I recommend more powerful models with larger databases like DeepSeek for a better experience.