Learning Chicken Cooking and AI Model Development
rswfire describes learning to cook chicken using his Ninja Foodie after previously being afraid of preparing raw chicken. He explains his progression from frozen chicken to raw chicken breasts, initially cooking them on his campfire using AI-provided instructions (tin foil on hot coals for 10 minutes each side), then discovering he could achieve the same results using the air crisp function on his Ninja Foodie.
He transitions to discussing his week-long effort to get a local AI model working to mirror him accurately. He has tested multiple inference engines including llama, BL LLM, and Xlama, encountering compilation errors. His goal is to have the local model process all his videos, classify them, and determine which should be public versus placed in his subscription service buckets.
rswfire explains his subscription service has been active since December but contains no videos yet. He plans to use one subscription level where supporters can access non-public content and enable comments for communication. He believes addressing a known, supportive audience will change how he frames his content, as his past 18 months with his public audience have been "less than ideal."
He describes the broader AI architecture plan involving reflections, clusters, and clusters of clusters. Cost is a limiting factor - processing his 700 videos for a small task cost $20 via API, and he estimates his full project would cost $500-2000. He acknowledges he's working under different circumstances than his "old life" and cannot fully explain his commitment to his current path, but trusts his choices are taking him where he needs to go.