Project Log: Naming Things, Building a Badminton AI Demo, and Remembering That Scope Is Everything
This week split into two very different tracks.
One side was still placement work: feature naming research, quick interviews, survey design, and trying to turn vague feedback into something the team could actually use.
The other side was a competition project: a badminton social and AI analysis app that I built with a teammate. That one was public enough to talk about, which felt oddly refreshing after weeks of writing around NDA shaped holes.
The badminton project started with a fairly simple question: after people finish playing, what do they actually want to keep?
Not another bloated sports platform. Not a chat app with payment and social features glued onto it for the sake of looking complete. We kept pulling the scope back to one core experience: finish a game, record the good moments, and leave with a small, positive memory of how you played.
That meant cutting quite a lot.
We focused the first version on instant matching, camera recording, highlight clipping, and a lightweight positive profile of the player. I liked that phrase, “positive profile”, because it stopped the AI from becoming another cold scoring machine. In badminton, especially amateur badminton, people are not always looking for brutal judgement. Sometimes they just want to see that one rally where they moved well, or that one smash that made the whole evening feel worth it.
Scope control was the real work. Every new idea sounded tempting for about ten seconds. Chat? Payment? Club management? Rankings? Sure, all useful. Also all dangerous. A demo version dies very quickly when it tries to become five products at once.
So we kept asking: does this help someone keep the memory of playing badminton?
If not, it went into the freezer.
Back in placement work, I spent a lot of time on feature naming research for a drone related function. I cannot share the exact feature details, but the work itself was interesting. We were trying to find a short Chinese name that felt clear, natural, and not too technical.
Naming sounds like a small task until you actually test it.
A name that makes sense inside a product team can be completely flat to a user. A name that sounds clever can create the wrong expectation. A name that is technically accurate can feel like reading a settings menu. We ran quick interviews and a short survey, then adjusted the questions when people were clearly interpreting things differently from what we expected.
That was the useful part. The misunderstanding was data.
I used to think user research was mainly about collecting answers. This week reminded me that it is also about watching where the question itself fails. If people hesitate, ask what a word means, or choose an option for a reason you did not expect, the problem may not be the user. It may be the framing.
I also joined an AI workflow sharing session inside the company. The examples were very practical: meeting note automation, small data processing tools, agent based coding workflows, and different ways teams are using AI to reduce repetitive work. It was not the glamorous “AI will change everything” version. More like: here is a boring task, here is a small tool that saves an hour, here is where it breaks.
Honestly, I prefer that version.
The whole week made me think about AI as a practical assistant rather than a magic layer. In the badminton project, AI helped shape a more playful sports memory. In the workflow session, AI helped with internal productivity. In naming research, AI was less important than real human confusion, which is also a useful reminder.
Not every problem needs a model. Some problems need twenty people to look at a name and say, “I have no idea what this means.”
By the end of the week, our badminton project had made it into the final round. That felt good. Slightly unreal, but good.
I think what I learned this week is that making a product smaller is not the same as making it weaker. Sometimes the smaller version is the only version with a clear pulse.
A focused demo.
A name people can understand.
A survey that asks one thing properly.
A feature that does not pretend to be a whole platform.
That is the difficult bit, really. Not adding more. Knowing what to leave out.
Leave a Reply