Daily Learnings: Tue, Jun 24, 2025
Chaos and Order are not enemies, only opposites. — Richard Garriott
MOAR EXPERIMENTING
I’ve build some additional things with Claude Code, and continue my experimenting with a smaller agent-based AI coding approach.
So far the paid $20/mo Anthropic plan is holding out, but I’m about to have Claude run on a large, multi-sub-agent task that will likely bump up against some limits. I won’t divulge much about what that is here, as it’s a bit of a SOLVD offering that’s proprietary IP. But I’m very interested to see how it runs.
Follow Up
I put CC through its paces tonight, running a lot of large-context tasks that included generating some UIs for testing and inspiration. I honestly was really impressed with how much I was able to get through without hitting the limit. But, at the end of the generation, I saw the warning that I was approaching my limit, with the limits to reset in about 4hrs time.
That said, my tasks were single-agent in nature. If I experiment with heavy multi-agent work, I can see the $20/mo plan running into issues. At that point, the $100/mo plan might make more sense.
MCP Servers and Claude Code
I’ve set up MCP servers for other tools like Cline, but today I started registering some for use with Claude Code. Turns out is pretty simple. Here’s the command for getting the Playwright MCP server up and running:
claude mcp add playwright npx @playwright/mcp@latest
# Verify installation
claude mcp list
Important Note: I ran this command first while already in Claude Code in bash mode (hit the ! key to run arbitrary bash commands). After the initial install, I ran the /mcp slash command to see if it was registered, but it returned “No MCP servers found”.
Turns out you should do this before running Claude Code (or restart your session) to get the MCP servers up and running. I exited CC and then restarted it and we were good to go.