Practical writing about prompt engineering, agent systems, and building real things with Claude. No hype, just what works.
Long-form thinking on the tools, patterns, and systems that actually work.
The context file that turns Claude from a chatbot into a teammate. Here's how to write one that actually works.
Read → March 2026 Prompt EngineeringEvery AI feature ships with a hidden product decision: the system prompt. Most teams treat it as an afterthought. Here's why that's wrong.
Read → February 2026 MCPSupabase, Notion, Figma, Slack, GitHub — all connected through MCP servers. What happened when Claude could read and write to everything.
Read → February 2026 Agent SystemsThe leap from "AI chat" to "AI operations" requires rethinking everything. Agents need roles, tools, memory, and boundaries.
Read →Most people skip straight to complex chains. But a few good examples in your prompt can outperform elaborate scaffolding.
Read → January 2026 Claude CodePre-commit validation, auto-formatting, test runners — hooks turn Claude Code from a tool into a workflow.
Read → December 2025 MCPOne MCP server, three deprecated API integrations, and a much simpler architecture on the other side.
Read → December 2025 WorkflowsStructuring Notion databases so agents can actually find and use the information they need.
Read → November 2025 Case StudiesSharing the process, the failures, and the systems that actually ship. Building in public is the best accountability tool.
Read →New posts when something real ships. No schedule, no filler.