Why Smart Developers Use Kimi For Design

BBetter Stack
Computing/SoftwareVideo & Computer GamesInternet Technology

Transcript

00:00:00If you head over to Design Arena, sitting right at the top is not a model by Claude,
00:00:04OpenAI or Google, it's Kimmy. So today I wanted to test how true this really is.
00:00:10We'll run through five different demos, web design, 3D modelling, UI components,
00:00:15mobile app design and game dev. We'll compare Kimmy to models like Astra and Fable and I've
00:00:21been blown away by some of the results here. I think with a little bit of steering you can now
00:00:25produce production-ready work with these models, but I'll let you be the judge.
00:00:30We'll score each round fairly and then choose the winner at the end.
00:00:38Okay, so the first task was to design a landing page for a developer-focused product. In this case,
00:00:44it's a job queue for a Postgres database. The first run here is from Claude. The font choice,
00:00:49a little bit odd, I'm not going to lie. The animation here is really good though. You can see
00:00:52items moving between lanes in the queue. That was a really nice touch. We've got the install button
00:00:57here. Overall, the design looks pretty good, although I would say the font is a little bit
00:01:01weird. We have nice shadows here, decent examples of the different features of the product. All of
00:01:05these graphics are really well done. Overall, I'd say Claude did a pretty good job there. Let's look at
00:01:09OpenAI. Slightly better use of fonts, although the graphics don't look nearly as good. If we scroll
00:01:14back up to the Claude version, you can see we've got an animated version here, which looks really nice.
00:01:20OpenAI did static. Overall, the graphics as well don't look nearly as polished. They look pretty
00:01:26crap to be honest. Look at this chart here compared to the chart on Claude's side. You can see Claude
00:01:35did a much better job there. Overall, OpenAI's attempt not quite as good. Now, if we look at Kimmy,
00:01:42personally, I think this is the best one. We've got the animated graphic on the side,
00:01:46much better use of fonts. This, to me, already looks like something I could just deploy and I'd be happy
00:01:51with. A nice underline lives in Postgres here. I love the UX here for the NPM install. The graphics
00:01:56also look really nice, on par with Claude, I'd say. The whole page looks really balanced. And then
00:02:02finally, a call to action at the bottom. Hopefully, you would agree. I think Kimmy has definitely won
00:02:06this round. Okay, the next task was to generate a 3D model of a floating island that animates between a day
00:02:12and night cycle. This is Claude's attempt, so it looks okay. You can see there's some slight weird
00:02:17clipping on the waterfall here. If we switch between the day and night cycle, it looks okay. Although
00:02:22everything does look a little bit blocky. If we look at OpenAI, this is from Astra. So much better.
00:02:30You know, look how sort of weird the trees look on Claude's version. Very blocky. Astra did a much
00:02:36better attempt at this. The waterfall is still clipped, although overall, it looks much nicer. The whole
00:02:41composition looks better to me. We can see if we go between the day and night cycle, the lighting is
00:02:45also way less harsh. And then if we look at Kimmy, super blocky here. The waterfall is super clipped.
00:02:52Overall, I'd say this is probably the worst result of the three. So for 3D design, Kimmy probably
00:02:57doesn't hold up too well. I'd say the winner here is definitely Astra. And if you're enjoying this,
00:03:03guys, you do me a massive favor by subscribing to the channel. Now, let's get back to the video.
00:03:07The next task was to design a component library. We've got Cairn from Claude,
00:03:12Forma from OpenAI, and then Meridian from Kimmy. Now, this one is a really difficult one to judge
00:03:17because I think they all look pretty good. So we've got some button components here. You can see we've
00:03:21got inputs, super while balanced, and we've got some nice focus elements there. It looks really nice.
00:03:26If we look at the version from OpenAI, buttons look really nice with the icons. We've got some inputs
00:03:32here as well with focus elements. It looks really nice. We've got things like check boxes and radios.
00:03:37And then we've got the version from Kimmy. Again, buttons look really good. The inputs look super
00:03:41nice. Focus element as well. Slightly more of a glow style there. We've also got radio buttons,
00:03:46check boxes. So overall, I think this is super equal. I'd find it difficult to say which exactly is the
00:03:52best one. Probably comes more down to taste. So I'm going to give a point to all three here.
00:03:57The next task was to design a mobile app for a coffee rating application. This was the worst
00:04:02round in my opinion. I think every model actually did a pretty poor attempt at this. This is the
00:04:06version from Claude. It looks super generic and basic and I would never be happy to release something
00:04:11like this. Very ugly design overall. We've got the version from OpenAI. It's a slight improvement,
00:04:17but again, super blocky and generic. I feel like with these sorts of requests, one shot in it is not
00:04:23going to be easily possible. You probably have to refine this over many, many prompts. And again,
00:04:28we've got the version from Kimmy. Super similar to the OpenAI version. Blocky, very basic and generic.
00:04:35Overall, I wouldn't be happy to use any of these. So unfortunately, because they are all equally
00:04:41terrible, they're all going to get no points. The final challenge was to generate a game and design
00:04:46all of the assets for it from scratch as well. So we've ended up with this robot jumping game.
00:04:50This is the version from Claude. So the model looks okay. It's pretty good detail on the eye. You can
00:04:55see it also blinks there as well. And the animation is pretty nice. We can also jump. And then we've
00:05:00got this flame animation, which looks pretty good. The walking animation also looks really nice. And
00:05:06we've got the smoke coming out the back. If we look at the version from OpenAI, slightly uglier for
00:05:11sure. The animation doesn't work at all. It almost runs backwards. The flame also looks way more basic.
00:05:19And you can see we've got this animation of some smoke maybe shooting off the screen there. Looks
00:05:23overall pretty weird. I'd say this is a pretty poor result. And finally, we've got the version from
00:05:27Kimmy. Overall, I'd say it looks slightly better than the OpenAI version. The feet actually move in
00:05:33the correct direction. The eye looks okay. Actually follows the mouse. Slightly less polished than the
00:05:39Claude version. Flame looks okay. Much better than the OpenAI version. I'd say overall, Claude is much
00:05:45better. OpenAI is definitely the loser here. And then Kimmy comes in the middle. So the point for this round
00:05:50is going to go to Claude. So overall, it seems like each of the models definitely has its strengths and
00:05:55weaknesses. Kimmy seems very good at things like web design. Claude is much better at interactive demos and
00:06:01games. And then OpenAI really shines with 3D modeling. So depending on your requirements, it's probably best
00:06:08actually to switch between each of these three models to get the best results. Now I ran Astra and Fable
00:06:14through the subscription. So I wasn't sure exactly how much that cost. But Kimmy cost just $6 to run
00:06:19all five demos. And if you want to see how well Kimmy K3 performs at other software engineering tasks,
00:06:24we've got a video for that right here.

Key Takeaway

Kimi excels at web design and costs six dollars for five demos, while model strengths vary across tasks with Claude leading in games and Astra winning in 3D modeling.

Highlights

  • Running all five design demos on Kimi costs a total of six dollars.

  • Kimi delivers the strongest landing page design for a developer Postgres job queue compared to Claude and OpenAI.

  • Astra outperforms Kimi and Claude in 3D modeling by producing a less blocky floating island with less harsh lighting during day and night cycles.

  • All three models perform poorly on mobile app design for a coffee rating application, resulting in zero points awarded for that round.

  • Claude wins the game development round by producing a robot jumping game with better eye tracking, working animations, and flame effects.

Timeline

Landing Page Design Evaluation

  • Claude produces a landing page with good animations and clear product features.
  • OpenAI generates static graphics that lack polish compared to Claude.
  • Kimi wins the web design round with balanced layout, custom fonts, and an animated graphic.

The first task focuses on building a landing page for a developer tool involving a Postgres job queue. Claude includes moving items between lanes and good shadows, though font choices are unusual. OpenAI provides a static and unpolished version with inferior charts. Kimi integrates an animated graphic, superior font usage, and a deployment-ready layout.

3D Modeling and Day-Night Cycle

  • Claude creates a blocky floating island with clipping waterfall issues.
  • Astra provides a better composition with softer lighting transitions between day and night.
  • Kimi delivers the weakest 3D result with severe blockiness and clipping.

The second task tests 3D modeling capabilities by generating a floating island that transitions between day and night cycles. Claude struggles with blocky tree assets and waterfall clipping. Astra, running through OpenAI's integration, resolves the lighting harshly and presents a much cleaner composition. Kimi falls short with the lowest quality execution in this category.

Component Library and Mobile App Design

  • Claude, OpenAI, and Kimi tie on the component library task due to equally high-quality buttons and inputs.
  • All models fail to deliver a functional or attractive mobile app design for a coffee rating application in a single prompt.
  • Zero points are awarded for the mobile app round because every output looks blocky, generic, and unpolished.

The third and fourth tasks evaluate UI components and mobile application interfaces. The component library features well-balanced buttons and interactive focus elements across all three models, resulting in a shared point. Conversely, the coffee rating mobile app round yields universally poor results that require extensive multi-prompt refinement.

Game Development and Model Comparison Summary

  • Claude wins the game development challenge by generating a functional robot jumping game with smooth animations.
  • OpenAI produces a broken walking animation and basic flame assets.
  • Kimi costs six dollars to execute all five demos and serves best in web design tasks.

The final challenge requires generating a complete game and all corresponding assets from scratch. Claude takes the win with detailed eye-tracking and proper flame and walking animations. OpenAI fails with backward-moving animations, while Kimi places in the middle. The overall workflow benefits from switching between models depending on whether the requirement is web design, games, or 3D modeling.

Community Posts

No posts yet. Be the first to write about this video!

Write about this video