diff --git a/portfolio/index.md b/portfolio/index.md
index 7f4d0f9..3e0d877 100644
--- a/portfolio/index.md
+++ b/portfolio/index.md
@@ -12,9 +12,11 @@ kernelspec:
 ---
 
 (portolioindex)=
-# Earning Level 3 
+# Deepening your knowledge
 
+Completing the basic assignments that demonstrate you can apply what we cover in class to your own data, is enough for a B, to earn an A you need to extend your knowledge. 
 
+You have some different pathways to earning an A.  
 
 ```{code-cell} ipython3
 :tags: [remove-input]
@@ -65,8 +67,50 @@ portfolio_cols = [ 'Level 3'] + ['P' + str(i) for i in range(1,5)]
 portfolio_df = rubric_df[portfolio_cols]
 ```
 
+## Extending Assignments
 
 
-Starting in week 3 it is recommended that you spend some time each week working on extensions.
+Starting in week 3 it is recommended that you spend some time each week working on extensions to earn level 3 on the skills.
 
 Use the feedback you get on assignments to inspire your extensions.
+
+To submit these, submit the work to a separate `<assignment>extended` branch so for assignment 2 extension, submit to `assignment2extended`.
+
+You should **not** extend *every* assignment since skills overlap and relate to one another. 
+
+I recommend making an issue that is a plan and asking for feedback before you work on extensions, so that you do not do too much extra work. 
+
+Some ideas:
+- extend A5 to add in a database source or combine the data in new ways, plus do extended EDA to earn level 3 for: access, construct, prepare, visualize and summarize
+- extend one of A7-9 to have experiments that more carefully analyze and compare different models at a task for one of the ml tasks (classification, regression, clustering) and visualize and summarize
+
+## Deployment and Distribution
+
+```{warning}
+this is currently a draft, will be updated by 10/24
+```
+
+Instead of earning all 15 level 3s you can earn any 10 plus the following: 
+- transform your portfolio to a publish-able portfolio by making it a jupyter book (mostly set up already, just need to clean up)
+- share a model and use a classmate's model with [huggingface](https://huggingface.co/CSC310-fall24) (making the model card would count for process and fitting it could count for one of the 3 tasks)
+
+if you go this route, I recommend level 3 for: 
+- access (loading models from huggingface will count)
+- python 
+- process (see [](process.md))
+- summarize
+- visualize
+- evaluate
+- one of: classification, regression, clustering
+- optimize
+- compare
+- workflow
+
+If your dataset of interest is hard to work with, images, or text, you might swap in 
+representation instead of access
+
+this means skipping level 3 for:
+- prepare
+- construct
+- 2 of:  classification, regression, clustering
+- representation
\ No newline at end of file