A book you would like to write
Book: The Differential Topology of Loop Spaces
Why: Because they are one of the first examples of spaces that are almost, but not quite, entirely unlike manifolds. They are relatively straightforward spaces which can be fairly conceptually grasped, but still contain enough intricacies to reveal some of the important differences between finite and infinite dimensions (though perhaps I should say between manifolds modelled on Banach spaces and more general manifolds). A book on their differential topology would thus be a gentle introduction to the topic than is (as far as I'm aware) currently available (in particular, although just about everything I'd want to say is covered in Kriegl and Michor's works, it's in such a context and with such generality that "daunting" doesn't quite cut the mustard).
Who For: Me, 10 years ago. That is, I'd try to write the book I wish I'd had when starting out in infinite dimensional differential topology so I wouldn't have made all the mistakes that I made.
Why Me: Because I work in that area and I think I've made just about every wrong assumption about loop spaces possible so I know lots of the traps for unwary differential topologists venturing out into the miasma that is infinite dimensional topology.
Will I Ever Actually Write It: Maybe, maybe not (vote for this answer if you want me to!). I made a start by writing up some seminar notes. I've started transferring them in to the nLab (but in the process I've been generalising them which slightly goes against the purpose of the project as I described it above). I'd certainly like to write it, if only to convince myself that I no longer have all those false assumptions, but whether or not I ever actually do it ... (hey, I've an idea, maybe all the time I put into MO and meta.MO could be reallocated to book-writing. Then it'll be finished next week.).
Update: 2019-01-07 Due to changes in circumstances, I am extremely unlikely to spontaneously develop the above-mentioned notes into a book. Should anyone be in a position to say to me "If you did polish those notes into a book we'd definitely publish it" feel free to get in touch.
Question seems a little silly to me, unless it's meant as motivation. But for those who answer the question and then are motivated to go ahead with their book project, I can offer some personal experience on the process.
Step 1. Start with a detailed outline and 100+ pages of detailed notes from a course that you've taught on the subject.
Step 2. Estimate about how long you think it will take to turn those notes into a published book. (In my case, I figured that it couldn't take more than a year or so.)
Step 3. Triple the value in Step 2 to arrive at an accurate estimate.
While I find the question borderline, I succumb to the temptation to answer.
Knot Theory: Kawaii examples for topological machines.
Topology is full of big machines, which may seem rather daunting to the student. But knot theory is a wonderful playground for toy models of many of these machines, where you can see how they work and visualize what they are doing. And one can draw pictures.
I think that a collection of these examples would be useful to students (I would have loved to have had it) or to people who would like to teach topology. And I don't think anything like this exists, really. The machine itself would be introduced only briefly, refering to somewhere else for more detail, while the knot theory example would be fleshed out in full.
For example, curvature of knots is the perfect playground for the Gauss-Bonnet Theorem. Computations of homology in knot theory give perfect toy examples (with pictures you can draw) for Mayer-Vietoris, the snake lemma, and other homological arguments. Ideas such as localization and Brown representability come up naturally. And an Alexander module gives a perfect playground for commutative algebra over a UFD.
So the idea would be to give sophisticated proofs of simple facts, letting the topological machines play the lead role. The student of topological machine X might then read the book by looking up the relevant section, which would give a kawaii (cute?) example in knot theory, highlighting how exactly the machine is working, and shedding light on its nature.
How likely am I to write it? I've toyed with the idea for a long time. For the book to be useful, it needs to be very visual and pedagogical, to make it light fluffy reading for one who knows the machine, and educational reading for one who doesn't. And becaue I have high asprations for it, it may take a while. But I do have intentions of actually writing it at some point, even if I don't yet know when that might be.