Blog
Daily experiences with AI, apps, development, and whatever else comes to mind.
One WhatsApp, Five Claudes: Multi-Session Routing and Monitor as a Message Bus
I wanted to run five Claude Code terminals from a single WhatsApp chat on my phone. Three limits broke in sequence — WhatsApp's 4-device cap, Claude Code's channel notifications not waking idle sessions, and no way to spawn new terminals remotely. Here's how each one got solved, and the one Claude Code tool I ended up using as a message bus.
We Don't Need AGI to Lose
Mythos was preview-released last week. It doesn't want anything. It can already bring down a Fortune 100 company. Why the AGI threshold is the wrong frame, from someone who ships agent systems for a living.
Two Lab-Grown Brains, One Game of Pong, Right in Your Browser
Real human cortical neurons, recorded by Cortical Labs, decoded into Pong paddles. A browser game played by biology.
I Spent an Evening Bisecting Anthropic's Claude Code Fingerprint
Setup tokens are supposed to bill Pro/Max plans. So why does Anthropic reject the same token from OpenClaw with a 'third-party app' billing wall? I went looking, byte by byte.
Nine AI Agents, One Shared Decision Problem
What happens when you wire nine LLM agents into a corporate hierarchy and ask them to converge on the same decision? An evening's experiment in multi-agent orchestration on a $5 sandbox.
Why I'm Running AI Agents in Slack Before Giving Them a Ticket System
AI agents need to talk before they can work. Here's why Slack channels with OpenClaw agents might be the better starting point than any ticket system.
I ignored social media for 15 years. That made sense. Then it didn't.
15 years of coding, zero posts, zero regrets — until the game changed.
Servus, World
First post — why I'm starting this blog and what to expect.