diff --git a/doc/presentation.org b/doc/presentation.org
new file mode 100644
index 0000000..78ce7d9
--- /dev/null
+++ b/doc/presentation.org
@@ -0,0 +1,53 @@
+* Introduction
+Write a static analyzer for ruby.
+
+Used an existing parser for ruby called `ruby-parse`.
+
+`ruby-parse` can generate an AST in JSON format.
+
+This made the task of writing a ruby parser simpler: Haskell comes
+with a great JSON parser.  So we just have to specify how to convert
+from the AST defined by `ruby-parse` to our own AST.
+
+* Implementation
+** ADTs
+*** Purpose
+Defining an Abstract Syntax Tree (AST) for Ruby using Algebraic Data
+Types (ADTs).
+
+*** Example
+    -- Value represents any JSON data (object, string, number, null, array)
+    Args       = Value
+    Name       = Value
+    Block      = [Definition]
+    Definition = Module | Function | Send
+    Namespace  = [Name]
+    Send       = {args : Args, namespace : Namespace, name :: Name}
+    Module     = {name : Name, block : Block}
+    Function   = {name : Name, args : Args, block :: Block}
+
+*** Live coding
+
+Define `Function`
+** Higher-order types
+Missing a good example here.
+
+** Type classes
+*** Purpose
+Be able to concert unstructured JSON data into our ADT.
+
+*** Example
+Consider the class
+
+   class FromJSON a where
+     parseJSON :: Value -> Parser a
+
+It's saying that `a` is convertible from json if there is a function
+`parseJSON` that takes a (json) `Value` and returns a `Parser a`.
+`Parser a` can be thought of as a procedure for generating a value of
+type `a`.
+
+So for instance.  The monomorphic type `Parser Function` is a
+procedure that generates a value of type `Function`:
+*** Live coding
+Define an instance for `FromJSON` for `Function`.