diff --git a/current_goal.txt b/current_goal.txt
index 60fa495..6d6b3ae 100644
--- a/current_goal.txt
+++ b/current_goal.txt
@@ -5,7 +5,7 @@ sitting in a tavern by a fireplace
 
 
 
-
+Use Machine Learning to build system to convert sentence -> behavior tree input
 
 
 
diff --git a/docs/design_ideas b/docs/design_ideas
new file mode 100644
index 0000000..de4cf40
--- /dev/null
+++ b/docs/design_ideas
@@ -0,0 +1,9 @@
+
+
+notes
+
+
+Create a neural net to generate human readable rules
+These human readable rules are they fed into an expert system
+This allows us to combine the learning ability of a neural net with the legibility of an expert system
+
diff --git a/docs/game/mechanics_goals.txt b/docs/game/mechanics_goals.txt
index 3140b8b..d2b65e3 100644
--- a/docs/game/mechanics_goals.txt
+++ b/docs/game/mechanics_goals.txt
@@ -2,11 +2,12 @@
 
 
  - Agents that you can talk to in the world
+ - Agents can execute a defined sequence of atomic actions
+ - Agents can learn new sequences based on natural language processing
  - Agents have task sequences that they are responsible for which they pursue
  - Agents can move from location to location
  - Agents can pick up items
  - Agents can place items
  - Agents can use items
- - Agents can 
 
 
diff --git a/docs/justification/feature_justification.txt b/docs/justification/feature_justification.txt
index bc3ef34..d62e730 100644
--- a/docs/justification/feature_justification.txt
+++ b/docs/justification/feature_justification.txt
@@ -20,3 +20,8 @@ This lets you not constantly scan a huge repertoire of information when you want
 Thus, I want to create an information heirarchy to mimic this and make engineering potentially easier.
 
 
+Generic Action Sequences
+In order to be able to command the ai to do something, have to have some way to represent chains of atomic actions for the ai to persue.
+Want to build complexity over time
+
+