On the AT01’s driver, the positive end of the LED is wired directly to the positive and of the battery/spring. The negative end of the LED goes through the µC to switch it on’n’off, and the other end of the switch goes to parallelled ballast resistors to lower the current.

So all I did was bypass the &micro’C’s switch so that it’d always be on, but still limited through the resistors, otherwise it’d be direct-drive. Probably wouldn’t kill the LED when using 3×AAA, but why bother?