Case Study
My AI operating system. It runs my professional life from a terminal. Not a prototype. The system I use every day across two jobs and multiple side projects.
An AI-powered operating system on top of Claude Code that manages my day-to-day across two jobs and multiple side projects. It reads my calendar, triages my email across 3 accounts, preps me for meetings by pulling contact files and past conversations, tracks decisions with 30-day review cycles, and manages tasks autonomously.
It runs 24/7 through 7 cron jobs: an hourly inbox manager that drafts replies (never sends), an hourly task worker that executes the highest-priority item, an 8:45am morning briefing, a 9am decision review flag, a 10pm end-of-day capture, and a Sunday morning weekly review.
The AI backbone. 165 skills covering marketing, SEO, design, code review, debugging, TDD, and workflow automation. 99 specialized agents for everything from security review to architecture planning.
Slack, Gmail, Google Calendar, Notion, Attio CRM, Apollo, Granola, Obsidian, Playwright, Excel, TurboVault, Context7, n8n, and more. Each one gives Claude Code direct read/write access to a different system.
11 memory files that persist across sessions: who I am, who I work with, my preferences, active projects, decisions made, tools used, design style, mistakes to avoid, references, and sales context.
7 cron jobs running on schedule. n8n workflows for multi-step automations. Decision tracking with automatic 30-day review flags. Task prioritization with autonomous execution.
Morning briefing generates automatically: today's calendar, email triage results, active tasks, goal progress, and a focus recommendation.
Decision review runs. Any decisions made 30+ days ago get flagged for retrospective. Did the expected outcome happen?
Meeting prep pulls the contact file, recent emails, CRM data, past meeting transcripts (Granola), and generates a structured briefing with talking points.
Inbox manager triages 3 email accounts. Task worker picks up the highest-priority item and executes it.
End-of-day capture: progress review, ideas captured, carry-overs identified, vault updated.
Weekly review: patterns, energy, goal progress, learnings. What worked, what didn't, what to change.
Not a chatbot. An orchestration layer connecting 21 systems through MCP servers, all controlled from a single terminal. Persistent memory means it knows who I talked to last week and what we agreed on. It drafts responses in my voice. It knows my preferences, my patterns, my mistakes.
Most people use AI as a tool they go to. This comes to me: proactively surfacing what matters, flagging what I'd miss, handling the operational overhead that used to eat 30% of my week.