Common Lisp. I basically only use SBCL. It has good introspection, restarts, and source analysis for debugging. I mainly write theoretical research code that doesn't depend on calling into the JVM or C++ code. I do try to keep my code portable, so I will check with other implementations from time to time.
I use GNU Emacs and Sly (though I am thinking of trying Lem). I don't use any structural editing outside of Emacs' built in electric-pair-mode, show-paren-mode, and expand-region (not built in). I don't even use rainbow delimiters anymore. I get all my Common Lisp dependencies from GNU Guix. It is very pleasant to use and is rolling release. In addition to Guix, I use cl-guix-utils, which adds live loading of dependencies quicklisp style.
I first learned Racket then Emacs Lisp (both in college). Emacs lisp was more pleasant due to its interactive and self documenting nature. I wanted to write real programs; Common Lisp looked and felt more like Emacs Lisp (but better). I started learning Common Lisp primarily with the "Lisp for the web" series. I was hooked. I learned more mainly through reading the hyperspec, studying other people source code and reading articles. I didn't read any of the famous books until I recently read "Practical Common Lisp". I already knew pretty much everything it had to offer. I wish I had read it sooner.