A circuit board is a bet you can't take back once it's etched in copper. Before we commit ours, there's one signal we have to trust, and it's the hardest one on the board.
Last time, Karl wrote about the call where Ryan and Kai asked whether a stopwatch could do what Slingshot does. The product survived that question. The answer was a single clean number a coach can read in seconds and a parent can believe in. But settling what to build only moves the problem down the bench. Now someone has to make a board that actually captures it, on a real arm, in a real gym, while a kid is going full speed.
The whole product lives in a pocket on an upper-arm compression sleeve. The board is smaller than a matchbook. Everything we want to know about how hard a player worked has to come from sensors packed into that space while the arm jumps, sprints, cuts, and lands. Nothing about that is a forgiving home for electronics.
The signal we lose sleep over is heart rate. Reading a pulse off a fingertip held still is easy. That's the demo everyone has seen. Reading it off a moving arm mid-session is the hardest signal on the board. Motion fights the optical sensor. The upper arm isn't the cleanest place to find a pulse to begin with. And basketball is nonstop motion, so the exact condition that makes the reading hardest is the condition we need it in.
Heart rate is the intensity context behind the Session Load Score. It's directional, a sense of how hard a player is working, not a medical readout. If that signal doesn't survive on the arm during live play, we don't tweak it. We move the sensor, change the optics, maybe rethink the placement. That's a redesign. It's the kind of thing you want to find out as early as possible, before you've built the whole board around it.
The part we trust, and the part we'll prove
The motion side, we're more sure of. Jumps, sprints, the explosive efforts that make up a session come from motion sensing fused with a barometer that reads tiny changes in altitude. Put together, a jump has a signature: the takeoff, the air, the landing. That's the part of Session Load we trust the design to capture. The board logs it all to onboard memory during practice, then dumps to a laptop over USB-C afterward. No app, no cloud, no live dashboard for v1. Capture, store, report. Simple on purpose.
The discipline that matters most before you order a board you can't change: build in a way to see what's alive. Test points on the power rails. A way to talk to each sensor one at a time. The unglamorous parts that turn a dead board from a mystery into a checklist. The first time we power this on, we want to ask "what's the smallest thing I can prove is working," and actually be able to answer.
One decision we made early pays off long after the pilot: the firmware is one codebase with two ways out. Today the data goes to memory and dumps over USB-C. Later, the same captured session can stream over Bluetooth to a phone, without rewriting the firmware. The hardware doesn't change between the pilot and the app. Only the way the data leaves the pod does. Designing that once, correctly, is cheaper than finding out the hard way.
We designed the whole board around a bet we can't fully settle until we hold it. That's the honest part of hardware. You commit, then you find out. Next chapter: we commit.