diff --git a/source/talks/extending-cruby-with-native-graph-data-type/notes/aeon-notes.md b/source/talks/extending-cruby-with-native-graph-data-type/notes/aeon-notes.md
new file mode 100644
index 0000000..77febd1
--- /dev/null
+++ b/source/talks/extending-cruby-with-native-graph-data-type/notes/aeon-notes.md
@@ -0,0 +1,67 @@
+---
+author: Anton Stroganov (@aeon)
+---
+
+Extending CRuby for fun and profit - Andy Pliszka
+- put your high level abstractions in ruby
+- put your low-level performance-intensive stuff in C
+
+- get ruby source from github and compile it (need openssl)
+- make test, make install...
+- install gems
+
+- run with lldb to debug!
+> lldb ~/my/ruby upstring.rb
+
+example: bignum.c
+- first, look for Init function
+- contains all the method definitions for the Integer, Fixnum class...
+- investigate the implementations of the methods to see how it all works
+
+How do you define a class in c?
+- define the method Init_Array()
+- rb_cArray is VALUE, which is basically object
+- rb_cArray = rb_define_class("Array", rb_cObject); # defines the array class and puts it in rb_cArray
+
+How to define a method on the class?
+- rb_define_method(rb_cArray, "length", rb_ary_length, 0); # arguments are - class to implement method on, method name, c function implementing the method, number of arguments the method takes)
+
+Two worlds...
+- C
+	- working directly with memory
+	- malloc/free
+- Ruby
+	- working with heap and objects
+	- GC handles things for you magically
+- switching between the worlds requires to make sure that you convert the values correctly when passing them between ruby methods to c calls and vice versa
+- use macros that come with the Ruby codebase (like LONG2NUM)
+- converting C functions to CRuby functions mainly requires changing the signature, input, and output to return VALUE (objects) instead of long or int or whatever.
+
+Why would you do this?
+- performance...
+- in a naive benchmark CRuby implementation of fibonacci formula is ~30x faster than plain ruby
+- a lot of methods are already implemented in C - numeric algorithm reference books, whitepapers, etc.
+- instead of trying to implement them in ruby, it can be worthwhile to just convert them to CRuby and compile them into the language itself.
+
+CLongArray:
+- init unwraps the ruby object to extract the pointer to the C array...
+- good reference to see how it accesses the struct with [] method
+
+CLongArray qsort:
+- basically a wrapper function that unwraps the ruby objects and....
+- passes the unwrapped value list to the dumb C quicksort implementation that has no idea about Ruby at all
+
+Example: Graph representation
+- simple Adjacency-List representation
+- C data structure containing an array of linked lists representing nodes.
+- ... code code BFS code code...
+- magic! 19x faster than plain ruby on a 1M node graph.
+
+Pitfalls:
+- try to avoid holding references to ruby objects in c code so they don't get garbage collected from under you
+- there is options for Data_Make_Struct that let you mark objects as used so garbage collector doesn't kill them 
+
+Do you really need to compile it into ruby? 
+- No, just write a ruby extension - you might want the ruby source just for reference to see how things are put together.
+
+- http://www.rubyinside.com/how-to-create-a-ruby-extension-in-c-in-under-5-minutes-100.html
\ No newline at end of file