Building Maestro: What I Learned Automating My Mornings with AI
How I built a multi-agent system that pulls your calendar, weather, Strava data, and priorities while you sleep — and why the hardest part wasn't the AI.
AIMulti-AgentPythonOpen Source
How I built a multi-agent system that pulls your calendar, weather, Strava data, and priorities while you sleep — and why the hardest part wasn't the AI.
A candid look at the real-world challenges of running Retrieval-Augmented Generation in production — chunking strategies, embedding model drift, and vector database gotchas.
How I built a system that automatically traces bugs to the commits that introduced them — turning 'who broke this?' from a blame game into an automated answer.
Traditional CI/CD pipelines assume deterministic outputs. Here's how I adapted my pipeline practices for LLM-powered applications where 'correct' is fuzzy.
What airline booking mistakes teach us about prompting LLMs. If you don't give them a map to Australia, they might send you to Austria.
AI doesn't need 'please.' It doesn't need context about your feelings. But your coworkers do. Let's not lose our humanity while learning to talk to machines.
Why Model Context Protocol is shifting LLMs from models that 'know' things to agents that 'do' things — and what that means for the future of AI engineering.