LAN Party Blog

Claude helped me develop and then ran some deep research prompts focused on domains that are important to LAN Party: TV drama, narrative game mechanics, audience engagement, location design, agent design, and prompt engineering. I turned each research artifact into a project subagent using this template, tailored for each subagent by Claude.

When agents are running in my console, I love to see basic functionality fail on ominously truncated strings being eval’ed in Python via Bash tool calls for some reason.

Claude Code is supposed to know what agents are available and when to use them, but it calls them more frequently if you add encouragement somewhere.

ActiveModel validation errors make great feedback for LLMs.