https://asciinema.org/a/1267653 shows my #uxn in #ilo assembly running the drifloon assembler, building another copy (properly byte identical) of drifloon, then assembling opctest & tictactoe, and running both.
The uxn implementation is down to 2853 cells, of which 2k are used for the stacks & internal memory buffer for block i/o, so just over 800 cells of actual machine code. It can be built to either write stderr into the blocks or to just discard the stderr text (which is what I have set as the default).
I'll do a little more work on this tomorrow to tighten up a few things, but it's complete. Later this year I will be working on a #nga port, and will work on expanding the device coverage, but I'm pretty happy with it at this point.