I think LW could do very well to import a lot of knowledge from NLP and try to see what's valid and what's not. I've noticed that people here are often reinventing the wheel.

Except when the wheel is, you know, 'bad'. Lesswrong wouldn't accept NLP for normative reasons. That said, I would be reluctant to give NLP 'wheel' status in any case. "Reinvent the swiss army knife" perhaps. :)