Decision tables, by the way, _work_. If you haven't tried them, it will only take a brief perusal of Wikipedia to get the concept. Their benefit is that all logic is represented in a single table: if every entry in the table has been filled, all cases are covered. Much easier to read than linear code, much less prone to error.
You can write your own decision table by hand in a spreadsheet program. But above a certain level of complexity(the point where you really, really want correct logic the first time) a complete decision table becomes very time-consuming to write by hand. By design it will still give more correct results than just hacking away, but automated approaches become very appealing. Decision tables were researched for a while in the 60s and early 70s, but were abandoned presumably as the field moved on to more prestigious topics.
Subtext's so-called "schematic table" does much to add the necessary automation - the next step in making it production-worthy would be to port such a system to output code in existing languages.
Look, this is a gimmick. Not having used such a tool, I'm not going to say that it's worthless. But I will say that I'm very suspicious of anyone who looks to gimmicks as the Salvation of Modern Programming. I've seen it too many times.
That being said, I've always been wary of "visual" tools that promise the sky and the moon, and I think I would go nuts if I had to use the mouse like in the demo, but it definitely looks interesting.
I do still have a fondness of flow-based programming, as is used in Kamaelia, twisted flow, or graphical VHDL environments though.
If throughout the video presentation you were wondering, like me, if it was mouse-only, the FAQ says:
"Subtext 2 has been designed to be much less mouse-centric than the prior version. Although not yet implemented, it should be possible to do everything from the keyboard."