AI Automation System - Setup, Model Allocation & Key Learnings
Original AI Mastermind pitch deck - an overview of the complete AI automation system. Slide 1: The Setup Two AI agents working in tandem: - CASE (Mac Mini, CTO/CMO): Content creation, LinkedIn outreach, social media automation, infrastructure - TARS (MacBook, COO/CRO): Operations, client pipeline, revenue tracking, meeting prep Shared data via git sync. Primary: MiniMax M2.5. Fallback: Sonnet 4.6. Slide 2: Feedback Loop Architecture Core concept: Data Collect > Analyze Patterns > Rewrite Strategy > Execute (loop). 7 active feedback loops auto-optimize outreach, content, engagement, and sales conversion. Slide 3: Daily Schedule Morning crons, afternoon engagement, evening content scheduling, overnight batch processing. Slide 4: Model Allocation Tiers - Premium (Opus): Strategic rewrites that control other crons - Standard (Sonnet): Human-facing tasks, DMs, judgment calls - Budget (Qwen/MiniMax): Mechanical tasks, templates, clicking buttons Slide 5: Before vs After Before: Manual everything, 0 pipeline, no systems. After: Automated outreach (20 connections/day), content pipeline, CRM tracking, automated follow-ups. Slide 6: Key Learnings - Sub-agents need explicit completion criteria - Ralph loops prevent context overflow - Browser profile isolation is critical - Model tier selection saves significant cost - Feedback loops require data logging from day one Slide 7: Full Tech Stack OpenClaw, Claude, MiniMax, Qwen 3.5, Facebook Graph API, Instagram API, Upload-Post, Calendly, Himalaya email, git sync, bash scripts. Slide deck available - reply if you want the link.