Vibe Pie Chart: Exploring Different Models - review
The Wrap Up
Let me share my final thoughts on this little experiment
I'm going to grade these pie charts and try to include them all one page because I did ask that they be modular. If the goal was to create a production quality pie chart I would have to put them in this order, with a 5 being reasonable for production and 0 being unacceptable.
Copilot - 4.0 Overall
Code: 3
It followed the D3 template and made some easy to read decisions. It leaves a lot to be desired for modularity, but it does the job in this case.
Design: 5
I wasn't really planning on doing a design portion for this because I didn't really give any prompting for this aspect, but there was such a wide range that it feels necessary. They really took that clean elegant Microsoft styling and applied it to this tool. I guess this may be subjective, but I think that it's unassuming based on my no information.
Functionality: 5
It has a button the left that changes the values in 5 categories. Nicely done.
Compliance: 2
It only really came through on about 50% of the things that I asked for. This wasn't too complex of an example in my thinking. I really expected it to have looked at something that exists out in the wild for this. I mean, there's a D3 example that is still overall better instrumentation of this. This area really hurt Copilot in my mind.
Qwen3 Coder - 3.5 Overall
Okay, so adding them all to one page is a pipedream. I've just included and image of what Qwen3-coder was able to accomplish.
Code: 2
The code was a super junior implementation. It was all linear with tons of repeating and just barelling through requirements and stepping on its own feet.
Design: 4
I'd give this a 4 in that it meant my requirements and took just enough liberty to make something that wasn't awful looking
Functionality: 4
Exactly what I asked for, nothing more, nothing less. It's not winning any prizes, but I'm not fighting it. The ding on this front is the animations don't work. I've been through the code a couple of times to find the bug and it had two conflicting animations one that was finishing right away and one that was to run for a second. These end up just being a flat transition.
Compliance: 4
It's all there. From the outside looking in this was a perfect execution. When you look under the hood it kind of falls apart, especially since the code and styles ened up not being modular even though I asked for that specifically in a seperate request.
Claude Code - 2.5 Overall
Code: 4
Nice clean code, but it wasn't modular like I asked. This needed to be in a class or contained in scope somehow. Instead it would break other code on the page.
Design: 2
This overall looks like a professional tool until you start moving the mouse around. It feels broken and clunky with unnecessary animations and awkward design elements. It's not responsive and it falls apart with different datasets.
Functionality: 4
Again, it does deliver on what I asked for. The poor design really hurts functionality, and the animation is not what was described
Compliance: 2
It got a lot right. Don't get me wrong. On the other hand, it made a ton of assumptions and went way off from what I asked in other cases. I just needed a simple modular pie chart that displayed different datasets. And I got something that feels really incomplete because it's trying too hard.
Meta.AI - 1 Overall
I don't even know if it's worth going over this one.
Code: 1
Buggy and novice junior. I was not impressed.
Design: 2
It does have the elements to be fixed up into what I asked for and what I expected, but it's a long ways off. Not responsive.
Functionality: 2
Again, the bugs and the lack of animations.
Compliance: 2
It did less than the bare minimum to get this working. It was obvious when I was copying the code that it was incomplete. There just wasn't enough there.
Conclusion
This was fun. I have to say that I learned a lot about how AI codes and it's capabilities and it's shortcomings. I use AI in my coding environment and if the change is more than a couple of lines at a time I don't really trust it. This is a clear demonstration of that in action. There was a lot to be desired still, but it did (with varied results) meet an impressive goal of using D3 to create a plugin. Maybe most impressive was the dummy data it generated.