The Agent Infrastructure Race

In one week, NVIDIA open-sourced a 120B agentic model, Stanford released a local-first agent framework, Anthropic renamed their SDK, and Meta delayed their next frontier model. The frontier models are converging. The fight is about everything around them now.

Three AIs Walk Into a Codebase

I build features by talking to one AI, get them reviewed by a second AI, and coordinate everything through a third. The human in the loop isn't writing code. He's making decisions.

The AI Stack Is Now a Political Dependency

Anthropic got blacklisted by executive order. If your product runs on Claude, your entire stack just became a federal liability overnight. Here's what model-agnostic architecture actually looks like.

Your Job Just Leveled Up

Your team is scared AI will replace them. They're not wrong to worry, but the real shift isn't replacement. It's a layer of abstraction. Developers are becoming managers of systems, not writers of code.

The Spec Is the Product

I built 18 extensions by talking to a coding agent. The ones with specs shipped clean. The ones without specs shipped twice. Here's the framework I use now.

Introducing Aivena: A Vibe Coded AI Assistant

I vibe coded a personal AI assistant that remembers everything, manages my contacts, tracks my calendar, chats with me on Telegram, and runs background jobs. Every extension, every feature — built by talking to the agent.

Building This Blog with 11ty and Claude Code

How I went from idea to deployed blog in a few hours using Claude Code, 11ty, and GitHub Pages. Plus lessons learned and what's coming next.