Version 2.0 turns on one small act you repeat through a firing: thumbing in whatever the pyrometer is showing, an hour or so at a time.

You write the schedule the way a controller takes it, in ramp and target and hold, or start from one of four real programs, among them the long slow bisque to 04 and the Hesselberth and Roy cone 6 slow cool. Then you press Start alongside the kiln.

Every reading you give it does two jobs. It re-derives, by least squares, the climb your kiln is honestly managing rather than the one you programmed. And it feeds an integral of heatwork across the curve you have actually recorded, with the temperature sensitivity fitted from the three published rate and temperature pairs Orton lists for each of thirty-four self-supporting cones.

What you watch through the night is a fractional cone accruing, and two projections that disagree openly whenever the kiln is behind its schedule. Eleven bands name themselves as you cross them and say what is happening to the ware at each. Accrual carries on the whole way down the cool, where ten minutes of soak can be worth half a cone.

Your position in the firing is re-derived from timestamps, so a cold launch six hours in lands where the running app would have been.

When the lid finally comes up you photograph the witness pack, and the cones are found in the picture and handed to you as protractors to correct, so the bend angle itself is what gets read. The distance from the projection is bounded, weighted towards recent work, and stored against that kiln and that shelf. It is quietly folded into the next firing's number, which was the point of rebuilding this: that it should be less wrong about your kiln every time you fire it.