Most teams reach for "an AI" without picking which kind. There are really four useful patterns — vector search, retrieval-augmented generation, multi-agent systems, and training your own model — each with very different cost, complexity, and payoff. This article walks through when each one is the right tool, with examples from real systems including the pet portal product search.
A practical mental model for building tools and applications with LLMs. Treat agents less like APIs and more like motivated junior collaborators — give them context, constrain the problem, and design tight feedback loops. Lessons from working alongside an AI-native startup founder on real products.
Building multi-agent systems teaches the value of balancing complexity with efficiency. Simplicity, effective use of natural language, and clear examples are key to unlocking their full potential. This article shares practical lessons and highlights the role of natural language and examples in guiding agents effectively.
Infrastructure management used to be complex and time-consuming, often requiring deep technical knowledge and hours of troubleshooting. With AI, it’s now much easier and faster. This article outlines a three-step process for leveraging AI in infrastructure: setting up resources, debugging with AI digests, and iterating for optimization. This approach has made managing infrastructure more accessible, empowering more people to build efficiently and focus on strategic goals.