diff --git a/doc/README.md b/doc/README.md
index 6a7a51e..288737d 100644
--- a/doc/README.md
+++ b/doc/README.md
@@ -3,10 +3,11 @@
 ## Assembly guide
 
 * [Writing the glue code](glue-code.md)
-* [Running Collapse OS on an emulated RC2014](emulate.md)
+* [Running Collapse OS on an emulator](emulate.md)
 
 ## User guide
 
 * [The shell](shell.md)
 * [Load code in RAM and run it](load-run-code.md)
 * [Using block devices](blockdev.md)
+* [Using the filesystem](fs.md)
diff --git a/doc/fs.md b/doc/fs.md
new file mode 100644
index 0000000..4e613bd
--- /dev/null
+++ b/doc/fs.md
@@ -0,0 +1,68 @@
+# Using the filesystem
+
+The Collapse OS filesystem (CFS) is a very simple FS that aims at implementation
+simplicity first. It is not efficient or featureful, but allows you to get
+play around with the concept of files so that you can conveniently run programs
+targeting named blocks of data with in storage.
+
+The filesystem sits on a block device and there can only be one active
+filesystem at once.
+
+Files are represented by adjacent blocks of `0x100` bytes with `0x20` bytes of
+metadata on the first block. That metadata tells the location of the next block
+which allows for block iteration.
+
+To create a file, you must allocate blocks to it and these blocks can't be
+grown (you have to delete the file and re-allocate it). When allocating new
+files, Collapse OS tries to reuse blocks from deleted files if it can.
+
+Once "mounted" (turned on with `fson`), you can list files, allocate new files
+with `fnew`, mark files as deleted with `fdel` and, more importantly, open files
+with `fopn`.
+
+Opened files are accessed a independent block devices. It's the glue code that
+decides how many file handles we'll support and to which block device ID each
+file handle will be assigned.
+
+For example, you could have a system with three block devices, one for ACIA and
+one for a SD card and one for a file handle. You would mount the filesystem on
+block device `1` (the SD card), then open a file on handle `0` with `fopn 0
+filename`. You would then do `bsel 2` to select your third block device which
+is mapped to the file you've just opened.
+
+## Trying it in the emulator
+
+The shell emulator in `tools/emul/shell` is geared for filesystem usage. If you
+look at `shell_.asm`, you'll see that there are 4 block devices: one for
+console, one for fake storage (`fsdev`) and two file handles (we call them
+`stdout` and `stdin`, but both are read/write in this context).
+
+The fake device `fsdev` is hooked to the host system through the `cfspack`
+utility. Then the emulated shell is started, it checks for the existence of a
+`cfsin` directory and, if it exists, it packs its content into a CFS blob and
+shoves it into its `fsdev` storage.
+
+To, to try it out, do this:
+
+    $ mkdir cfsin
+    $ echo "Hello!" > cfsin/foo
+    $ echo "Goodbye!" > cfsin/bar
+    $ ./shell
+
+The shell, upon startup, automatically calls `fson` targeting block device `1`,
+so it's ready to use:
+
+    > fls
+    foo
+    bar
+    > mptr 9000
+    9000
+    > fopn 0 foo
+    > bsel 2
+    > load 5
+    > peek 5
+    656C6C6F21
+    > fdel bar
+    > fls
+    foo
+    >