Daily Learnings: Mon, Jun 30, 2025
If I find 10,000 ways something won’t work, I haven’t failed. I am not discouraged, because every wrong attempt discarded is another step forward. — Thomas Edison
Follow up on Claude Code Experimentation
Following up on my experience with Claude Code, and breaking up tasks into smaller chunks with fewer assumptions: Things are moving a lot smoother. I have had the agent reviewing each prototype HTML page, comparing it to the existing site by reviewing code and using the Puppeteer MCP server, and then proposing a plan to update the prototype to fit more with the current site’s content. As I iterate on this, I’ve developed a new slash command to speed up the process. I’ve also asked Claude to run it’s plans by Gemini using the Zen MCP server, which has yielded some really interesting adjustments and suggestions to the designs.
That last piece is what’s been the most interesting about this process. Instead of it just being a simple “change this text” series of tasks, it’s actually been continued adjustments and iterations on the designs themselves, landing the prototypes in a better place in terms of data presentation and navigation.
Some quick learnings to document:
- Zen MCP has been helpful in creative, open-ended tasks
- Sticking with pure HTML, CSS, and JS for prototypes keeps it lightweight enough for Claude to manipulate things quickly and output high-quality designs
- I’m still really surprised at how much I like using Claude for design work. It’s taken a bit to dial it in, but it’s really upped my game in terms of value that I can offer as someone that has never really been a designer.
- Slash commands continue to save me time and effort when working on semi-repeated processes
- They very-well might be my favorite thing about working with Claude Code. ESPECIALLY when compared to working with Gemini’s new CLI, which I played with last week (which doesn’t have a concept of custom slash commands yet)