diff --git a/notes.txt b/notes.txt
deleted file mode 100644
index ac42667..0000000
--- a/notes.txt
+++ /dev/null
@@ -1,31 +0,0 @@
-* Context
-  * Community protocol
-    * Hydration timing doesn't really work
-    * No way to await future provided context
-    * Should we define global state to represent pending context requests?
-  * New implementation
-    * Requires everyone to agree on the single implementation.
-    * Imperative API means it can be used outside Hydrator.
-  * Hydrating timing alternative
-    * Could run `@provide()` immediately but delay `hydrate()` by a microtask.
-    * This would mean that reading context is done asynchronously from the initial
-      execution which gives an opportunity for ancestor components to hydrate and
-      provide that context.
-    * This would make `getContext()` more ergonomic in `hydrate()`, since we would be
-      reasonably sure any ancestors are already providing the relevant context.
-    * Doesn't work in practice because it would make hydration logic circular. For
-      `@provide()` to execute first, it needs to also get the value for that property,
-      which likely comes from an `@hydrate()`. To run `@hydrate()`, we need to hydrate
-      descendant components to support syntax like
-      `@hydrate('child-component', (el) => el.count)`.
-    * Hydrating children first means they will request context before it can be provided
-      by the parent.
-    * Even if this wasn't an issue, we can't generally assume that all ancestor
-      components will hydrate synchronously, since we can't assume they are all
-      implemented with Hydrator.
-    * As a result, we still run into the same problem of requiring a descendant
-      component to wait for context to be provided by an ancestor component.
-    * This approach also breaks user-space `connectedCallback()` and
-      `disconnectedCallback()` because they can run before hydration and nothing is
-      initialized.
-    * TL;DR: This alternative just doesn't work.
\ No newline at end of file