AI Agents run my business and life

Key Concepts

• Vibe Coding: A methodology of building software and automating tasks by interacting with AI models (like Claude) in natural language, often bypassing traditional coding expertise.
• AI Agents: Autonomous or semi-autonomous software entities (e.g., OpenClaw, Harbor) that perform specific business or personal tasks by executing workflows, managing data, and making decisions.
• Vector Databases: Systems (e.g., Gbrain, Pinecone) that store vast amounts of data in a format searchable by LLMs, allowing AI to "remember" and retrieve context from personal or business documents.
• Context Window: The amount of information an AI can process at once; increasing this allows for more complex, long-term management of businesses and personal life.
• Agent Teams: A framework where multiple specialized AI agents (e.g., Dev, Marketing, Support) collaborate to solve complex problems, often outperforming single-agent setups.

1. AI-Driven Business Operations

Andrew Wilkinson utilizes AI agents to manage his holding company, Tiny, and his personal family office, Folly Partners.

• Administrative Automation: He uses agents to handle support tickets, accounting, and email triage. High-priority issues (P0) are automatically addressed by a "Dev Agent" that can fix code and merge pull requests.
• Marketing & Growth: An AI agent integrated with PostHog manages Meta and Reddit ad accounts, performing multivariate testing and adjusting budgets autonomously.
• Financial Oversight: His CFO built a custom dashboard using Claude that replaces expensive software like Adapar. It integrates bank, accounting, and market data to provide a real-time global balance sheet and portfolio stress testing.
• The "Eye of Sauron" Effect: By ingesting historical data into vector databases, Wilkinson can query his entire conglomerate to identify trends, accounting errors, or underperforming CEOs.

2. Personal Productivity & Life Management

Wilkinson treats his personal life as a product to be optimized through AI.

• Deep Personality: A startup he built using AI that administers psychological tests and generates a 100-page, Robert Greene-style report on the user’s archetypes, attachment styles, and career suitability.
• Personalized Media Diet: He uses an agent to scrape his newsletters and emails, then uses Gemini Voice to generate a custom, 7-minute daily podcast that only includes information relevant to his life, city, and business goals.
• Health Optimization: An agent named "Mara" tracks his Apple Health data (HRV, sleep, temperature). It successfully identified a correlation between his wrist temperature and the onset of viral nerve pain, providing predictive warnings.
• Life Context: Using an app called "Hearsay," he records his entire day to build transcripts, which are then ingested by Gbrain to provide his AI agents with full context of his daily interactions.

3. Methodologies & Frameworks

• The "Interview" Prompting Technique: Instead of writing complex prompts, Wilkinson instructs the AI to "interview" him for 5–10 minutes to gather all necessary information. This ensures the AI has the correct context to build a perfect output.
• Agent Teams: He advocates for using a "team of eight sub-agents" for any complex task. By assigning different personas (e.g., a team of various medical specialists for health queries), the quality of the output increases significantly.
• The "Cigar Butt" Investment Strategy: Wilkinson views the current software landscape as a "cigar butt" market—previously high-quality businesses are now being commoditized by AI. He suggests investing in foundational infrastructure like TSMC or data centers rather than betting on individual software startups.

4. Notable Quotes

• "I think we're probably three to six months out from being able to just hand businesses off to AI to run, at least basic businesses."
• "Software is a worse business now than it was five years ago... you are picking up pennies in front of the steamroller in most industries."
• "Never expect someone to understand something that their paycheck depends on them not understanding." (Regarding the difficulty of transitioning legacy employees to AI-first workflows).

5. Synthesis & Conclusion

The core takeaway is that the barrier to building complex software and managing large-scale operations has collapsed. Through "vibe coding" and agent-based architectures, individuals can now perform tasks that previously required entire departments. However, Wilkinson warns that this democratization creates intense pricing pressure and competition, turning many software businesses into "cigar butts." His strategy is to use AI to achieve extreme personal and operational efficiency while investing in the physical infrastructure (chips and data centers) that powers the AI revolution.