Thanks for the hints - I’ve added the template script and such, and updated my post. I found that the time calculation itself could be optimised, since the call to parse "%e" sys.now and the math that followed was taking a timestamp, turning it into an object, then turning it back into a timestamp again (but where the start of the epoch is midnight of the current day, BMT), and it could be shortened to just (1/86.4) * (86400 % (sys.now + 3600)). Matters very, very little when the speed of the Lil interpreter is limited within Decker, but it feels better to me.