Two-Week Check-in #5: What do you think of when you hear "jerk" 🤨

Get your head out of the gutter. It's ACTUALLY a physics concept (google it, I promise).

If you ever sat or drove a manual car 🏎️, when there is a bad shift, the car jerks you back and forth. What you're actually feeling is a sudden change in acceleration - that's jerk. And it's something we added to our gait detection algo.

Last update we told you we were jumping to the 32g accelerometer and I questioned the pricing. We're trying to solve both the measurement and the pricing on multiple fronts.

The 32g chip is expensive AF 💸

Like, expensive to the point where it's threatening our whole philosophy: affordable, and PER shoe. The spec sounds fine, the price doesn't.

Two things we're doing about it:

1. Designing a cheaper accelerometer onto the board. Still 32g, but less resolution, hopefully still good enough for what we need.
2. Switching the gait detection algo so it can flex across different accelerometers instead of being married to one.

And the experiment underneath all of it: can we get clean step detection at 16g? We still need 32g for the deeper metrics, but 16g is the hard-mode version of the problem. If the algo can nail steps while clipping at 16g, it only gets sharper with the extra headroom at 32g. That's exactly why we added jerk.

Let me get nerdy on how adding jerk affects our algo 🤓

Picture the accelerometer as a ruler that only reads up to 16 cm. Sprint hard enough and your foot blows past 16 cm, but the ruler just reads "16" the whole time it's maxed out. Since we'll never know what the max value actually was, our gait detection doesn't work.

We still use this max value, but now with "jerk" we also take into consideration the change from 0 to 16 cm. So when our accelerometer detects a sudden change from 0 to 16, our 2nd layer of gait detection comes through. It doesn't care how big the hit was, only that it happened suddenly. And a footstrike is the most sudden thing your foot does.

The result 📈

We're sitting at ~90% step detection accuracy for BOTH walking and running. That's progress - one front down 🔥

We have social media now 📱

Finally. It's run (GET IT?!) by a friend of ours who's growing his own page, and he's a genuinely dope creator. Go show him some love at @chun.mov.

We're on Instagram and Twitter/X, nowhere else for now. A few early posts are up.

Instagram: @mylz.run Twitter/X: @mylz_run

Go follow, and tell us what you actually want to see on there.

Keep logging mylz, ✌️ Brian