In partnership with
Hey folks,
Claude Code recently launched a mini version of Claude Design available by using the /design command so you can now invoke Designs within Claude Code.
I run an AI tools directory. Thousands of listings, each one a page telling you whether a tool is worth your time.
I have looked at those pages hundreds of times but have never bothered with a design overhaul.
This week I asked Claude for 4 different designs of one listing. 18 minutes later I had the redesign, and a list of everything wrong with the page that existed inside the Claude Code app, without having to do my work separately in a standalone app and try handover the designs.
Let's dive in..
In Claude Code
Here is the first prompt:
The only 3 things in there that mattered:
"Take a look at my AI tools listing website on [link]." I pointed it at the real website for context.
"Come up with some alternative canvases." Ask for one /design and you have to work out whether it is right. Ask for several and you just pick options.
"To decide whether that's a good AI tool fit for their use case." That is the job the page has to do. Notice what is not in the brief. No colours, no layout, no reference sites, nothing at all about how it should look.
Then nothing happened
13 minutes for the first cut. Another 5 finishing off in the background. 18 minutes before I had anything to look at.
I am not going to dress that up. Sitting there with no progress bar, no idea whether you are 2 minutes from done or 20, is genuinely unpleasant. Go and make a coffee.
It was not idle, mind you. 7 commands, and 6 separate trips out to my live website. It went and read the real page over and over before it designed anything.
What came back
4 directions that were betting on a different reason people leave the page.
1. Fit check first. A "is this you?" strip up front, good fit if, skip it if, before you start, with the verdict pulled up right behind it.
2. Show the product. Dark hero, one big screenshot, 3 captioned screens, and a CTA bar that stays with you down the page. Betting that nobody buys software they have not seen.
3. Spec sheet. Starts at, time to first result, learning curve, last verified, then one dense table. Betting the visitor has 5 tabs open and wants facts, not persuasion.
4. Comparison first. My tool sitting next to 3 alternatives, then a straight pick this if, pick something else if, do nothing yet. Betting they arrived comparing, so stop pretending otherwise.
3 of those genuinely surprised me. The first one did not, and that is the point. 4 options where every single one is brilliant means they were never really 4 options. I only needed one to be right.
The design audited itself
This is the part I would tell a friend about first.
Designing my page 4 different ways did not find a design problem. It found a content problem, and it found the same one 4 times over. Every layout came back with the same holes in it, just in different places:
The main image is the tool's logo.
You cannot see the actual product anywhere on my site."Freemium" is my entire pricing section. Not a single number.
The key features restate the tagline. One of them reads "Desktop app for Claude Code", which is not a feature.
The verdict, the one block that actually helps someone decide, sits at the very bottom under everything else.
"0 reviews" and "20 views" are both on display. Both of those actively hurt.
The similar tools are category matches, not alternatives.
I asked for a design and got an audit.
Then I started amending
Once the options were up I could click anywhere on the canvas to comment
Me: I like the navy blue header, however move this to the header of option C so I can see it in this design too
Add comment just leaves a note. Send to Claude makes the change. Same box, and you choose whether it is a reminder for later or an instruction for now.
/design
Claude Design is not new. It has been out since April and I use it frequently, for proposals, slide decks and general mockups. For that kind of work it is still the better tool by a mile, and it has features this new /design command does not come close to.
The design was never the problem.
I would spend an afternoon in Claude Design building the perfect landing page for an app idea. Really nail it. Then hand it over to Claude Code to build, and what came back was mostly built. Close enough that you cannot really complain, off enough that you keep noticing. Small inconsistencies, everywhere.
So then I am fixing design problems inside a coding tool, which is the worst place on earth to fix a design problem.
/design does not do better than Claude Design. It makes it simpler for Apps. What it does is never leave Claude Code. The design and the build now happen in one window, one conversation, one set of context. There is no translation step, so there is nothing to lose in translation.
If the design ends in a document, do it in the design tool. If the design ends in a working App, do it in Claude Code
Stop making AI decisions in the dark.
Leadership is asking: are we getting value from AI? Which tools are worth the spend? Where are we exposed? Right now, most teams have no idea.
Harmonic Security Usage Explorer changes that.
You get a complete picture of how your organization uses AI, automatically categorized into custom tasks and use cases.
You’ll see the projects being worked on, who’s using what tools, where AI investments are driving value, and where employees are engaging in risky behavior.
CIOs can rationalize spending and cut wasted licenses. CISOs can pinpoint where risk exists and neutralize it. AI committees can show exactly how their efforts are paying off.
👉️ Forward this to someone who needs to know about /design in Claude Code.
Forward and unlock the Vault.
Cheers, Tim
