This is a silly side project I built for my friends and I. You submit your body measurements and let AI figure out what sport you may have a competitive advantage in, given your dimensions.
How it came about: Some friends and I, who are all past our peak physical years, were chatting it up at the bar, and we got to talking about who among us would still win a race in the sport we played in high school, without any training. Naturally, our current body shape/dimensions became the primary factor in our conclusions. So it got me thinking, what dimensions are _most_ informative in our ability to excel in a sport, and, given these dimensions, what would an LLM have to say about it.
Having shared this among the friend group, the results have been super interesting. In some cases, like mine, the results were surprisingly on target with the sports I actually played in high school and have excelled at in my life. I swam in high school, and specifically free style and butterfly strokes, and this was the first sport, down to the stroke, that the AI chose for me. Second was surfing, which has been my lifelong passion, and something I did competitively in college.
For others, the results have been very unexpected, like a D1 volleyball player who had fencing picked as the top sport.
On the technical side, I'm happy to share some details about what went into this. I challenged myself to "vibe code" as much of this as possible in Typescript, mostly in the interest of seeing how far I could push these gen-AI tools, from design to implementation. This was hit or miss, and required a lot of hand holding to get it production ready. Where it completely fell flat was in the design and construction of the morphing avatar. Spatial awareness still seems like a bridge too far for AI coding tools, and the model had to be built with morphs manually (and frankly it still needs a lot of work). The rest of the stack is:
- Typescript
- SQLite
- OpenAI's GPT 5 model
- Server hosted on AWS EC2
- Frontend is a very basic Next.js page hosted on Vercel.
- WebGL and canvas for the avatar
Anyway, hope this brings a little fun to your day. Give it a shot and LMK your feedback.
This is a silly side project I built for my friends and I. You submit your body measurements and let AI figure out what sport you may have a competitive advantage in, given your dimensions.
How it came about: Some friends and I, who are all past our peak physical years, were chatting it up at the bar, and we got to talking about who among us would still win a race in the sport we played in high school, without any training. Naturally, our current body shape/dimensions became the primary factor in our conclusions. So it got me thinking, what dimensions are _most_ informative in our ability to excel in a sport, and, given these dimensions, what would an LLM have to say about it.
Having shared this among the friend group, the results have been super interesting. In some cases, like mine, the results were surprisingly on target with the sports I actually played in high school and have excelled at in my life. I swam in high school, and specifically free style and butterfly strokes, and this was the first sport, down to the stroke, that the AI chose for me. Second was surfing, which has been my lifelong passion, and something I did competitively in college.
For others, the results have been very unexpected, like a D1 volleyball player who had fencing picked as the top sport.
On the technical side, I'm happy to share some details about what went into this. I challenged myself to "vibe code" as much of this as possible in Typescript, mostly in the interest of seeing how far I could push these gen-AI tools, from design to implementation. This was hit or miss, and required a lot of hand holding to get it production ready. Where it completely fell flat was in the design and construction of the morphing avatar. Spatial awareness still seems like a bridge too far for AI coding tools, and the model had to be built with morphs manually (and frankly it still needs a lot of work). The rest of the stack is:
- Typescript - SQLite - OpenAI's GPT 5 model - Server hosted on AWS EC2 - Frontend is a very basic Next.js page hosted on Vercel. - WebGL and canvas for the avatar
Anyway, hope this brings a little fun to your day. Give it a shot and LMK your feedback.
URL: https://www.mybestsport.com
- Dylan