As a programmer, I am tempted to say "unless the project is actually a large program". "Large" is relative, of course.

Of course, I have seen LightTable before the comment on LW, and I tried to imagine applying it to any basically data-crunching (as oppposed to mostly UI) program. Visualising computation may look like a good idea. Unfortunately, at the level it is demonstrated in the demo, it is simple enough for anyone who even tries to write a big program to keep it in mind.

When you have multiple layers of abstraction and each of them has a reason to do non-trivial double loops (which is not that much if you can say what each level is doing and why) what we see in demo would become overcluttered. I am not sure whether LightTable demo will grow into a tool to make UI fine-tuning more comfortable or it will try to invent some approaches that work for back-ends and isolated data-crunching. In the former case it will stay a niche thing but may become a well-olished narrow-focus tool. In the latter case it will have to transform so much that it is hard to tell whether the current developer will succeed.