The companion tool for your Small / Medium / Large robot build — works fully offline, no internet needed.
It's how many seconds it takes your robot's motorized arm to spin all the way around 5 times. Bigger robots usually have heavier arms, so the same motor takes longer to push them around — this gives you a second number (besides brick count) that should also change with robot size.
But this only works with certain motors — check yours first. There are two kinds:
Alternative metric — "load until stall": if your motor compensates for speed, instead measure how many extra bricks you can hang on the arm before the motor can't hold the target speed anymore and stalls out. That forces the motor past its compensation limit, so the number reflects the robot's real strength/weight again. Pick which metric you're using below — it changes the labels for you.
Connect directly to a LEGO Education Single Motor (from the Computer Science & AI kits) over Bluetooth, then use Start/Stop below to time a motor run automatically — no stopwatch needed. This also works great for the pre-check test above, and it can even auto-detect the exact moment the motor stalls.
https:// or from localhost — opening the file directly (file://) will usually block Bluetooth for security reasons. If the button below does nothing, try running a simple local server (ask a teacher/adult, or search "how to open a local file with a local server") and open the page that way instead. Make sure your Single Motor is charged, powered on, and broadcasting. This connects using the same open protocol as LEGO's own LEGO Education Python API.
Before pressing any buttons — look at your three dots and guess. What do you think the brick count and motor time will be?
Build a robot the exact same height as your Medium robot, but hide extra bricks inside it (or make it hollow). Log it here, then flip the switch below and press "Let the computer find the line" again.
The computer used regression — the same thing you did by eye, but calculated precisely — to draw the line that best fits all your data points. That's what "training" an AI often means: finding the line (or curve) that makes the smallest total error across every example it's shown.
When you added the Trick Bot, the line shifted — sometimes a lot from just one weird example. That's exactly how real AI models can get fooled: they assume new cases will follow the same pattern as their training examples. When something breaks that assumption — like a house that doesn't look like the houses a pricing model was trained on — the prediction can be way off with total confidence.