Looks like a very neat tool, it certainly has promise. Perhaps the best part of it is that it doesn't try to contain itself to tracing one kind of information (however useful that may be).
Full tracing inside an application is still dependent on uprobes, but for many use cases I think it would suffice to be able to read profiling (gprof) data and combine it with the kernel level information. This would really make it a powerful profiling solution. And with uprobes one has the best of both worlds: basic profiling of any program and very detailed profiling when using special profiling builds from the same tool.
But in practice the most used feature I have for strace would be the live view on an already running process. The "why does this program appear to be stuck" kind or "why is this program so extremely busy. And part of these questions is answered by the speed the output of strace scrolls by (especially when combined with a few filters to filter out some common event-loop "noise").
Seeing possible burst patterns (and immediately looking back at their names by suspending the view) is really very different from trying to interpret timestamps and creating a mental picture of the flow. Sure it's not very scientific but often it gives enough information about how to proceed. So live information feeds is (for me at least) one the main reasons for grabbing a tool like strace.
But I guess something similar will be created for 'trace' like what was done for 'perf' and things will all be good. Great work guys! :-D