The improvement for me:

* speed of typing : I really appreciate not having to take my finger off of the main part of the keyboard to use arrows or the mouse, I also don't like the 2 stack (2 mode) switching for vi. Somehow I find it easier to remember the few keyboard cords the toggling of modes.

* works in a terminal : launch with emacs -nw and you can work across an ssh terminal with same syntax highlighting, same shortcuts

* available for most OSes : my co-workers use fancy new editors except they have to rsync code from the server back to their latest Ubuntu, OSX or Windows just to edit it.

* people like to customize it and there are modes for all kinds of esoteric languages and build systems : i don't fiddle much with my .emacs so it is not benefiting me personally but others swear by it

1. On-the-fly keyboard macro generation. I use M-[, record, M-]. f9 to execute. (C-x e, e... if function keys not available) Even very small, one-time, but repetitive tasks can be automated.

2. First-class functions that can be bound to anything. I always bind "goto-line" to a single key (like f7). That alone saves me lots of time. And custom functions work the same way. You can put a macro to insert that if __name__ == '__main__' in python scripts with a single keypress, or with a "M-x <function_name>".

In addition to language boilerplate, you can use these facilities to help enforce arbitrary coding standards. I have a function template for C that adds a comment "/* end <function_name> */" after the close bracket, for example.

These facilities apply to any file you might want to work with. This is in addition to whatever editing modes might already exist for the language.

3. Working in text-only terminal on a remote server is only slightly different from working on my desktop. Not having to manually sync files or mess with X configuration is a huge win for me.

Emacs is not an IDE- its a text editor. It would be nuts to use Emacs to edit Java compared to Intellij or even Eclipse. Having said that, it is no more productive than any other text editor at editing text (as compared to Vim or Sublime Text 3). Even if that weren't true, would you really invest years into learning a new editor to make editing text a few percent faster? Programming is limited by thought, not typing (common pain here is solved by plugins anyway). The benefits of using emacs are as follows:

1. terminal based 2. key chords 3. (e)lisp 4. open source 5. stable 6. extensible

What's a "point and click IDE"? Eclipse and Visual Studio are "shortcut based". I've never seen a "point and click IDE" except maybe SQL Windows/Gupta/Centura.

1. Doing funky things with symbol under cursor. In past lives ive had documentation for things like assembly instructions, opengl apis etc displayed inside emacs.

2. Emacs macros.

3. Lack of hopping between terminal and editor.

For me it's less about productivity and more about comfort. When I first started using Emacs I felt like it was a lot easier to move the cursor around and manipulate text with shortcuts consisting of keys that were already near where my hands were on the keyboard.

I use vim now though, and I find that the keybindings feel much more natural, being based on the positions of the keys rather than mnemonic-based ones in Emacs (e.g. ctrl-n for "next line").

EDIT: Sorry, I realize this doesn't really directly respond to "IDE replacement" - honestly I don't write much code any more so some of the below doesn't apply to that. When I wrote code, the main benefit of emacs was that it abstracted me a bit from the nuances of different IDEs for different languages / platforms and was good at generalizing interactions with the compiler / errors / objects. A really good IDE would be better at many certain things than emacs, but emacs was marginally better all around. I used it for C, perl, and a little C++.

1) "the kitchen sink" thing, which is commonly cited as a drawback, is (to me) emacs most compelling feature. If this isn't compelling to someone, then a lot of the rest of emacs probably won't be, either.

There's a learning curve to doing everything in emacs. I've written some guides on it before to ease the transition to starting to use it (basically introducing a few very important basic concepts that introduce major and minor modes, keybindings, and how to investigate those and customize them), but when you have that core understanding down, when you realize that you can apply those same concepts to everything you do - it's incredibly, incredibly powerful and you understand why people want to live in emacs.

eshell or shell-mode for your shell interactions is amazing. searching through a shell buffer with emacs' regexes or incremental search is fantastic. Running your whole shell in that buffer and being able to run simple but useful commands like "occur" (essentially letting you grep through multiple preceding lines of output) is something I use all the time. The commands to jump up / down through the buffer to prior prompts in the shell history - all of that stuff is great.

Similarly, using dired (directory editor) as my shell browser is great. If I need to do a batch file rename, I could jump through some hoops to do this with shell script one-liners or something, but putting dired into editable mode and then treating the file names simply like text and using search & replace operations the same way I would on regular content in a file is a natural, fast, and powerful operation.

And of course all of these are the same commands and same environment as I have for editing text.

I don't do everything in emacs but I do as much as I can in it. And I didn't mention org-mode but I use it for 100% of my note taking and publishing to HTML - it's incredibly awesome.

Eclipse has emacs-style key bindings. So you can have the best of both worlds.