diff --git a/MarkdownLivePreview.py b/MarkdownLivePreview.py
index 54617e5..932f759 100644
--- a/MarkdownLivePreview.py
+++ b/MarkdownLivePreview.py
@@ -26,8 +26,8 @@ def get_resource(resource):
 
 
 def plugin_loaded():
-    resources["base64_loading_image"] = get_resource("loading.base64")
     resources["base64_404_image"] = get_resource("404.base64")
+    resources["base64_loading_image"] = get_resource("loading.base64")
     resources["stylesheet"] = get_resource("stylesheet.css")
 
 
@@ -222,7 +222,7 @@ def _update_preview(self, markdown_view):
 
         basepath = os.path.dirname(markdown_view.file_name())
         html = markdown2html(
-            markdown, basepath, partial(self._update_preview, markdown_view), resources
+            markdown, basepath, partial(self._update_preview, markdown_view), resources,
         )
 
         self.phantom_sets[markdown_view.id()].update(
diff --git a/README.md b/README.md
index 2efebad..84b6a30 100644
--- a/README.md
+++ b/README.md
@@ -23,5 +23,4 @@ in `MarkdownLivePreview.py` and `markdown2html.py` (GitHub only shows the top
 3. All your code should be formated by black.
 4. Send a PR!
 
-
-
+FIXME: add a git hook to format using black (can the git hook be added on github?)
\ No newline at end of file
diff --git a/markdown2html.py b/markdown2html.py
index 85a7014..928426a 100644
--- a/markdown2html.py
+++ b/markdown2html.py
@@ -1,11 +1,17 @@
-import copy
+""" Notice how this file is completely independent of sublime text
+
+I think it should be kept this way, just because it gives a bit more organisation,
+and makes it a lot easier to think about, and for anyone who would want to, test since
+markdown2html is just a pure function
+"""
+
 import os.path
 import concurrent.futures
 import urllib.request
 import base64
 import bs4
 
-from functools import lru_cache, partial
+from functools import partial
 
 from .lib.markdown2 import Markdown
 
@@ -91,35 +97,35 @@ def markdown2html(markdown, basepath, re_render, resources):
 
 
 def get_base64_image(path, re_render):
-    def callback(url, future):
-        # this is "safe" to do because callback is called in the same thread as
-        # add_done_callback:
+    """ Gets the base64 for the image (local and remote images). re_render is a
+    callback which is called when we finish loading an image from the internet
+    to trigger an update of the preview (the image will then be loaded from the cache)
+    """
+
+    def callback(path, future):
+        # altering image_cache is "safe" to do because callback is called in the same
+        # thread as add_done_callback:
         # > Added callables are called in the order that they were added and are always
         # > called in a thread belonging to the process that added them
         # > --- Python docs
-        images_cache[url] = future.result()
+        images_cache[path] = future.result()
         # we render, which means this function will be called again, but this time, we
         # will read from the cache
         re_render()
 
+    if path in images_cache:
+        return images_cache[path]
+
     if path.startswith("http://") or path.startswith("https://"):
-        if path in images_cache:
-            return images_cache[path]
         executor.submit(load_image, path).add_done_callback(partial(callback, path))
         raise LoadingError()
 
-    # FIXME: use some kind of cache for this as well, because it decodes on every
-    #        keystroke here...
     with open(path, "rb") as fp:
-        return "data:image/png;base64," + base64.b64encode(fp.read()).decode("utf-8")
+        image = "data:image/png;base64," + base64.b64encode(fp.read()).decode("utf-8")
+        images_cache[path] = image
+        return image
 
 
-# FIXME: wait what the hell? Why do I have two caches? (lru and images_cache)
-# FIXME: This is an in memory cache. 20 seems like a fair bit of images... Should it be
-#        bigger? Should the user be allowed to chose? There definitely should be a limit
-#        because we don't wanna use to much memory, we're a simple markdown preview plugin
-# NOTE: > The LRU feature performs best when maxsize is a power-of-two. --- python docs
-@lru_cache(maxsize=2 ** 4)
 def load_image(url):
     with urllib.request.urlopen(url, timeout=60) as conn:
         content_type = conn.info().get_content_type()