I love the new color palette, the clean conversation-style message/response display, and the ability to re-size the online friends box is clutch.
So far my only complaint is that I don’t like the icons without text in the control bar. Seeing the words right where I am clicking reassures me I am taking exactly the action I intend to take. I know I can hover and it will display the text but that takes additional time.
More importantly, it breaks the immersion. Historically inside of Gmail I have not needed to think about how I needed to accomplish the task of “doing email” whether that was reporting spam, archiving, labeling, or replying. Most software is a tool for accomplishing a greater task and the best software allows us to focus all our thoughts on the task without thinking about how we are using the software. The software developers removed the need to think about the specific tasks it takes to get the inbox opened, have the email displayed on screen, and create a new message to send as a reply or move the email to where I want to store it.
One example of this is a telephone. Think about when you make a call. You pick up the receiver, dial the number and talk to the other person(assuming they answer!). If the dialing of the number is the software interaction* then the key pad is the software. Imagine if every time you went to use a phone you had to concentrate on finding the numbers; that is the numbers were in random orders depending on the make, model, or version of the device. There is a chance you might lose track of either the number you needed to dial or what it was you were calling the person about because you had to shift your focus onto how to use the phone first. OK while that is a bit of a dramatic example, how many times have you been using the latest and greatest web app only to get lost in the documentation trying to figure out exactly how to use that new feature they promised you and then completely lost interest or worse forgot the details of what you wanted to do with the new feature? Your immersion in the task was broken because you had to think about the details of how to operate the tools needed to accomplish the task.
So how does this idea tie to the text-less icons? A lot of websites use very similar icons but with varying meaning associated depending on which website it is and, at least for this user, that creates uncertainty about what will happen when the button is clicked.
For example, what would you expect to happen when you click a button with this icon?

Archive? Maybe. But you probably said download. This is because it looks very similar to the icon used for most download buttons.
Email has become so critical in our lives. I do not want to have to worry if I am clicking the wrong button nor waste time waiting for a pre-action confirmation. By adding the words under the icons the user would be assured that they are clicking the archive button and not downloading a copy of the potentially confidential email to the downloads folder.
*technically hardware interaction but just go with it…