There is some confusion, you are agreeing with me.
I mean don't try solving novel problems in your day job with novel tools and techniques. Use known tools to solve problems. Lean new tools when not solving problems.
OK i misunderstood you, but I don't see how you can avoid learning new tools when solving problems.
You must often hit the point where you think "surely there is a better way to do this, lets research it" - then find new tools, spike them and then use them? No?
However what I think you are getting at maybe is people who think stuff like "Node.js - that sounds funky, oh.. Hadoop - thats hip at the moment how can we use that?" perhaps for fun, or to enhance their CV. That is dangerous. You end up with completely the wrong tool for the job!
Yes, that's what I mean. Software is all about building tools.
I mean like "I know, I'll do this client job in Lisp, I've never tried that" which is probably a bit of an extreme example but I have learned similar by experience :)
I mean don't try solving novel problems in your day job with novel tools and techniques. Use known tools to solve problems. Lean new tools when not solving problems.