AI
Large language models, agents, prompt engineering, and the practical side of shipping AI features.
- Thinking in agents: from single prompts to multi-step workflowsWhen a single LLM call is not enough, you reach for an agent. Here is the mental model I use to design them without overengineering.
- RAG in practice: what actually breaks at scaleBuilding a retrieval-augmented system is straightforward in a notebook. Keeping it honest in production is not.
- Building with Claude skills: a practical introAnthropic's Claude skills turn abstract 'prompts' into reusable, reviewable units of behavior. Here's how I've been shipping them at work without ceremony.