The video's still on my phone. I never watched it.
I started with the smallest problem I had.
Hi, I'm Sophie. I didn't pick up tennis until I was 30 — and for a long time I did the same thing every week: record the session, tell myself I'd review it later, never review it.
Not because I didn't care. Because the video was 58 minutes long, and maybe 36 of them were actual tennis. The rest was walking, picking up balls, adjusting my racket, standing around. I wasn't going to scrub through 22 minutes of that to find the rallies buried in between.
So I built something to cut them out. I prototyped it alone — wrote the code, tested on my own footage. It's ready in seconds, not minutes: a decision I made at 2am, because I wanted to watch my session before I forgot how it felt.
That's Volosophy Rally Clipping. It's free, forever, and it works today.
Then I realized clipping was just the first thing I could see.
Once I could actually watch my sessions, I noticed things I'd been blind to for years — the same mistake three weeks running, a serve that fell apart whenever the score got tight. Not subtle. I'd just never watched.
But seeing on its own goes nowhere. I'd notice the mistake Monday and lose it by Thursday. The video remembered the session; nothing remembered me — my patterns, what I'd fixed, what I hadn't. It reset every week.
And even if something remembered, I hit the next wall: I knew what I was doing wrong. I didn't know what to do about it. Something has to turn "here's what happened" into "here's what to work on next."
And say I knew what to practice — then what? Coaching is expensive and inconsistent. Hitting partners are hard to schedule. A ball machine gives you repetition, but it doesn't know you — it feeds you the same ball whether you've fixed that shot or not.
That's when it clicked. The problem was never any one of these tools. It's that nothing in the loop remembers you, and nothing hands off to the next thing.
You watch, but it goes nowhere. You practice, but it doesn't compound. Every piece works alone — so every session starts from zero.
So we're building the loop — one system, where each piece feeds the next. You see what happened. It gets remembered — your patterns, your rhythm. That becomes what to work on next. That becomes real practice, on a machine that feeds you the ball you actually need instead of the same one on repeat. Then you go again, with more clarity each time.
Miss one link and the chain breaks — which is exactly why this can't be a feature. It has to be a system.
And that's why rally clipping matters more than it looks. It isn't the product — it's where the loop starts. You can't understand what you never watched. Everything downstream depends on first being able to see. It's live today, free forever. The rest, we're building one piece at a time — each clearing the same bar clipping did: real, and working on a real court. We'll tell you when the next one's live. Not before.
And while we build the software, we're building the machine.
Right now there's a ball machine in our lab firing its 200,000th shot. Nobody's watching — it just runs, all night, so we can find out what breaks. Because there's no point knowing what to practice if nothing can actually feed you that ball. So we're building the thing that can — and testing it, one shot at a time, until it holds up on court and not just in a demo.
But let me be clear about what this machine is for. It's not here to replace the person across the net. I don't want a future where everyone practices alone against a robot and calls it tennis — that version of the sport makes me a little sad. A machine should make your practice sharper, so your time with real people is better. Not stand in for them.
Why I think we can build this
None of this is one product. The software has to work on real footage — bad light, wind, a phone propped against a fence. The hardware is a different animal: structure, electronics, firmware, motor control, tooling, a factory floor. Then the two have to talk to each other, on a court, in the sun.
Almost nobody has done the whole chain.
I've done all of it. I grew up on factory floors — my family builds things for a living. I'm a robotics hacker by training, with degrees from UC Berkeley and a CS master's from UPenn. I was the first product person at one of China's largest AR glasses companies, the only PM on a team of engineers, and took products from prototype to mass production. The AR startup I led product at went from prototype to unicorn.
And I actually play tennis. That sounds like the least important line here. It isn't. Plenty of people can build a robot. Very few have stood on a court wondering why the machine keeps feeding the same ball to a backhand they fixed three weeks ago.
I'm building the thing I needed.
Fast over fancy. Free over freemium. Clipping before coaching. Every decision so far came from my own frustration, tested on my own footage.
Everything we ship has to clear the same bar rally clipping did: it has to be real, and it has to work today. We'll tell you when the next piece is live. Not before.
Progress is a loop.
Observe. Practice. Adjust. Repeat — with more clarity each time.
Training should remember you.
Your history, patterns, and rhythm should shape what comes next — and that starts with being able to watch what you did.
Technology should deepen the game.
Not mechanize it. Done right, it disappears — and what's left is a game that feels more personal, more alive, more yours.
Now I'm joined by a team of builders bringing that idea to life.
If any of this resonates, we'd love to meet you — as a player, coach, builder, creator, tester, partner, or early believer shaping what comes next.
Come train with us. Come build with us.
— Sophie Luo, Founder of Volosophy
