Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Sure, until everybody hits the wall where everything from everyone looks the same no matter how well made it is.

Talented designers are not getting replaced.
Nope. They will just be using new tools.

But one thing I will say is that people with limited physical artistic talent (can see it in their mind but there’s a disconnect when trying to draw it) may finally get tools to let their visions be realized.
In the past you would need to find a person to try to articulate your vision, and often they would instead articulate their vision.
 
  • Like
Reactions: Populus
TL;DR: Useful but not a designer, better at code than deign, likely expensive use of tokens

So I tried Claude design for a project I am working on that needed an infographic for displaying survey results. I did not want a complicated one, just a racetrack style graphic.

I am by no means a graphic designer, so I gave it the areas to use and asked it to display the results in an infographic to see what it recommended; I wanted to see if it had a better idea than the racetrack style I wanted to use. Here are the results and timeline:

  1. After asking some questions about the purpose, orientation and size, Claude Design came up with the standard color coded tiled dashboard, like every results display ever created. Its code was solid, but the result generic.
  2. I prompted it to build a racetrack style display. It did, not bad, but looked like a first cut, draw on a napkin design. It had a number of design elements that I thought were a bit off - the corners were not used but just filled in black, a large center black infield with the legend and title, and the proportions looked wrong. In fairness, I specified portrait and A4 size so it proportioned the design accordingly.
  3. Now the fun started. The biggest issue I had was explaining what I wanted and getting Design to actually do it. It took multiple steps to get it to size the segments properly and to stop putting black vertical segments at each end of the graphic.
  4. Once I got it to only have the active, i.e. color coded based on results, segments, I had it adjust proportions so it looks right. That took several steps because it would interpret "make corners 1.25 in height art the rectangles" as instructions to make the rectangles larger, not the corners smaller. Eventually I got it to where I wanted it.
  5. Design added numbers to each segment, something I did not ask it to do and the corner numbers did not fit entirely in the segment on one corner because Design placed all in the same upper left location and did not account for rounded edges. I deleted the numbers.
  6. Then I asked to place icons in each segment representative of each topic. It did, and the placement was good but the icon designs looked like clip art from PowerPoint. I will be replacing them.
  7. Once I liked the results, I downloaded the code. It looked solid and I can modify it for my needs, adding variables for some of the elements to make it DRY.
Conclusion

  • Claude Design is good for prototyping but any production ready design will take work and need a human hand to finalize. The final output is something I could hand to our designer, say this is what we are looking for, and ask it to be improved.
  • The HTML/CSS code it generates for use looks solid, as does the svg code for the actually graphic. Rather than the approach I was using, which was to use an svg for the racetrack and variables in the svg code to color code, ad text, etc., it used a flexbox to generate the track. Much better approach than mine, and only used svgs for icons.
  • Design details such as icons look generic. ChatGPT did better one than Design.
  • It is very slow. I often thought it was frozen when it was actually working and would eventually give results. Part of this was a poor UI that hid the Claude pulsating icon to indicate it was working, weird scrolling behavior, and cryptic error messages.
  • It is likely expensive. My test used all the free tokens for the week.
Edit: added to TL;DR comment on coding ability
 
Last edited:
Exactly. People looking to design on cheap will get what they want pay for; if you aren’t willing to spend the time to think through your pitch , what does that say about your dedication to making things that are unique and great?



If anything they will. E more valuable and AI a tool they use to do some grunts work.
Just instruct the AI to take inspiration from a site and boom you'll get it. No designer has had an original idea either. It's just a faster way to do the same thing.
 
  • Like
Reactions: lapstags
Just instruct the AI to take inspiration from a site and boom you'll get it. No designer has had an original idea either. It's just a faster way to do the same thing.

I agree it is faster, but my experience with Claude Design is you don't get a finished design even when you prompt it to use specific designs as inspiration. It's good at mimicking but poor at thinking about what something should look like, or coming up with good ideas. It is a tool to that is good at rendering for putting someone's ideas on paper, as it were, that no doubt it will get better.
 
I guess we are heading to the era where UI/UX designers would be no longer needed. They will either have less job to do, a lot of work with prompts or no job at all.

I hope this thing can at least generate something better than the slop minimalist UI designs that these same designers were making for more than 15 years already. What’s in the picture is not really “tasteless” but kinda boring. Can it generate a Y2K inspired website with funny fonts, shadows and backgrounds?
I totally agree with you but it comes down senior leadership who don't understand design. They only understand money. Saving it and keeping it. Good enough is how they think, sadly. Unnerving times, for sure.
 
  • Like
Reactions: uacd
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.