Courtesy of Vladimir Jankijevic on the XSI mailing list, here’s how to accumulate values in a weightmap:

The nice thing about his tree is how it sets things into per-point context so nicely. My own attempt seems crude in comparison:

The thing about setting weight map values is that you can’t do it like this:

ICE will evaluate this tree just once, even if it is in the Simulation stack. Perhaps this is a case of mistaken optimization? (branch is a scalar constant, so no need to reeval?)

If you replace the scalar node with any of these branches, then the weightmap values will change as your play through the timeline.