Cracking Claude
Last Monday I had reserved most of the workday to begin experimenting with Claude and the new Figma MCP. Since then, I’ve had time (in fits and starts) to continue exploring and applying Claude to different areas of my daily work. I’ll keep this short, but if you’ve spent any time in relevant Reddit threads, you’ll see a lot of worry. I’m not going that direction; this post is focused on my own experience with three use cases so far. I’m eager to see where this all goes from here.
Token Test – convert fixed pixel dimension to variables/tokens
I knew what I wanted to achieve, did some reading and dove in. Through trial and error, Claude converted all of my fixed pixel dimensions to variables. That’s something like 56,000 individual properties adjusted. Along the way I save the skill for reuse. Days of what would have been carpal-tunnel-inducing clicks resolved.
My Voice – learn my writing style
I fed Claude a few of my own editorials and asked it to save my voice as a writing style. I added a few corrections to it (and will continue to do so) but out of the gate it was spot on to how I write. Like uncanny valley… With that tested I’ve started the same process for a work voice and expect to put that to the task for our next annual report.
Make this better – improve my Figma kit for handoff
My Figma work has been design focused – organizing information and processes and presenting them in an attractive way. My web design days go way back to Flash and early Dreamweaver. Before joining the Court, I’d been away from that type of digital work for nearly 20 years. So, to say I have a novel approach is a fancy way of saying newbie. With that said, I’ve been binging on UX and UI for since the end of ’24.
I know there’s huge gaps between my Figma kit and what a developer would really appreciate. While the design kit is moderately robust, it is missing a lot of things that I don’t even know about (being transparent here.) So, I asked Claude, “What would make this a better kit to hand off to a development team?” It came back with some positives and a list of improvements. Whoa. Ok, “Claude – let’s do that.” I reviewed the list and started with things that made sense, did some further reading/research, and had it make most of the recommended changes. These were mostly tedious tasks – adding descriptions, standardizing variable naming and properties, consolidating bits and pieces, etc. I had usage guides written for most of my components and it was able to reference that content for much of the changes. Changes that would have taken hours–if not days–just done!
The following day I came back and asked it to rerun the “…make this better…” prompt. The things it still flagged are items I’ll come around to strategically with further thought and some outside validation from our development team.
What’s next?
Build skills – both my own and Claude’s.