The 99.99% Threshold
I watched a couple videos on agentic coding recently:
The Name Ikigai
I wanted to share a bit about the name Ikigai.
Implementing /pin and /unpin
We recorded a full development session implementing the /pin and /unpin commands. Seven hours, unedited. This post walks through what happened and links to key moments in the video.
What Is Ikigai?
Ikigai is an agent orchestration platform we’re building. It will pair with the user, not with a project or a machine. It will sit above your projects as an umbrella you work through, aware of all of them.
Parallel Agents with Ralph
My typical workflow now involves 2-3 agents running in parallel terminals. One finishes, I feed it a new goal, and it’s off again. PRs get auto-merged. The cycle repeats.
New Tools For Ikigai
I just merged rel-09 into Ikigai. The headline feature was the web tools; search through brave or google plus a web fetch tool.
How Do You Trust AI Code?
People ask how I can trust code that an AI writes.
When --version Lies
Coverage checks were failing on GitHub Actions but passing locally. Both systems reported lcov 2.0-1.
Standardized Output for AI Agents
When you hand off a task to another person, you reduce their cognitive load by giving them clear, consistent information. The same principle applies to AI agents.
Stop Reading the Code
I told a coworker recently: “With AI, if you’re looking at the source code, you’re doing it wrong.”