diff --git a/.drone.yml b/.drone.yml
index 22e038c..20d0901 100644
--- a/.drone.yml
+++ b/.drone.yml
@@ -13,6 +13,7 @@ steps:
       - cd diagrams
       - ls | grep d2 | xargs -I{} d2 --layout=elk {}
       - mv *.svg '../images for report'
+      - cp *.svg '../report'
       - cd -
 
   - name: Build Project synopsis
@@ -28,11 +29,12 @@ steps:
   - name: Build Report
     commands:
       - cd report
-      - pdflatex report.tex
+      - pdflatex --shell-escape report.tex
       # Prepare bib
       - /usr/bin/vendor_perl/biber report
       # Compile twice for the table of contents and for bib text
-      - pdflatex report.tex
+      - pdflatex --shell-escape report.tex
+      - pdflatex --shell-escape report.tex
       - cd -
 #
 #  - name: Generate text
diff --git a/notes.txt b/notes.txt
deleted file mode 100644
index 7102a3e..0000000
--- a/notes.txt
+++ /dev/null
@@ -1,12 +0,0 @@
-Change title to Sky -> Sky UK Limited
-Say something about diferent devisions at Sky
-Add pictures of campus
-Change pictures to make readble
-Make sure the images have references
-Subscript images
-Make better distigisable bettween projects and case sutdies
-new pages bettween sections
-more details in the case studies less in the naratives
-in cases studies talked about problems that I had to over come
-
-https://www.youtube.com/watch?v=PKfR6bAXr-c
diff --git a/papers.txt b/papers.txt
new file mode 100644
index 0000000..fee62bf
--- /dev/null
+++ b/papers.txt
@@ -0,0 +1,27 @@
+Non papers, industry resources:
+- https://docs.aws.amazon.com/rekognition/latest/dg/what-is.html
+- https://www.clarifai.com/company/customers
+- https://cloud.google.com/vision
+
+- Model generation Base model generation
+	- Database search
+		- results for imagenet: https://paperswithcode.com/sota/image-classification-on-imagenet
+		- Comparasion of different classification methods https://ieeexplore.ieee.org/abstract/document/9687944
+		- EfficientNet: https://arxiv.org/pdf/1905.11946.pdf
+		- Imagenet dataset: https://ieeexplore.ieee.org/abstract/document/5206848
+	- AutoMl
+		- A survey of state of the art in ml - https://www.sciencedirect.com/science/article/pii/S0950705120307516
+		- AutoMl benchmark: https://arxiv.org/abs/1907.00909
+
+- Transfer learning / Chaging Output layer / Multi model approach
+	- https://journalofbigdata.springeropen.com/articles/10.1186/s40537-016-0043-6
+	- https://ieeexplore.ieee.org/abstract/document/9134370
+
+- Cost of machine training / Efficiency
+	- https://ieeexplore.ieee.org/abstract/document/8622396
+	- https://ieeexplore.ieee.org/abstract/document/9770050
+
+- Node Management
+	- Cache invalidation: https://dl.acm.org/doi/abs/10.1145/70082.68205
+	- Federation Learning: https://www.sciencedirect.com/science/article/abs/pii/S0950705121000381
+
diff --git a/report/report.tex b/report/report.tex
index 45a3d2d..9b41b96 100644
--- a/report/report.tex
+++ b/report/report.tex
@@ -122,7 +122,7 @@
 			\end{itemize}
 
 			% TODO add diagram!
-			% \includesvg{expandable_models}
+			\includesvg{expandable_models}
 
 		\subsection{Generation models}
 			The system requires the generation of models. Generating all models based on one single model would decrease the complexity of the system, but it would not guarantee success.