The geas disassembler should be usable not just for inspecting bytecode, but also for editing bytecode. One challenge with this are jumps. The disassembler output for jumps looks like thi:

     ...
     push 0x23
     jump
     ...
     jumpdest
     ...

There is no way to infer where the jump actually goes. To help with this, the disassembler can print PC numbers (geas -d -pc):

           ...
000a:      push 0x23
000c:      jump
           ...
0023:      jumpdest
           ...

This makes jumpdest offsets visible, but it also means the disassembler output is no longer a valid assembler input. But there is no reason it couldn't be. We just have to make numeric labels valid in the input language, and turn them into a kind of assertion. When the assembler encounters a numeric label, it should check the computed PC at that point matches the label.