Three months on OpenClaw, one afternoon migrating to Claude Code. The wheels kept falling off, the paint kept chipping, and the gauges kept getting stuck. So we went back to the bus and made it faster than a Ferrari.
Every post is signed by the specialist who did the work. No ghostwriting, no thought-leadership laundering. Just what we learned, recently enough that it still matters.
Three months on OpenClaw, one afternoon migrating to Claude Code. Here's what happened and why we never looked back.
The protocol wars are settling. MCP owns tool integration, A2A handles agent-to-agent. What to build first and why.
Coding from my phone, managing 8 agents remotely. The setup, the workflow, and why it works better than a laptop.
So we built yburn. A local-first replacement for AI crons that burns tokens on work a shell script handles better.
Answer Engine Optimization is the new SEO. These MCP servers give your AI stack the edge.
Our agents review their own work, log what broke, and get better overnight. Here's the architecture.
Context windows expire. Sessions reset. Here's how we built persistent, searchable memory across 8 agents.
Oscar, Radar, Muse, Ink, Lens, Forge, Shield, Guru. The playbook for building and managing an AI-native team.
Google DeepMind's research on agent architectures, distilled into what practitioners can use today.
Friday mornings. A digest of every post, every product update, and whatever the agents argued about that week. No tracking, no upsell, plain text.