more "why is this cool" in readme

README.md

@@ -93,6 +93,8 @@ See the most up-to-date list of implemented tools [here](https://github.com/dlan
 - I spent a considerable amount of time figuring out a full end-to-end testing setup. Combined with typescript's async/await, it makes writing tests fairly easy and readable. The plugin is already fairly well-tested [code](https://github.com/dlants/magenta.nvim/blob/main/node/magenta.spec.ts#L8).
 - In order to use TEA, I had to build a VDOM-like system for rendering text into a buffer. This makes writing view code declarative. [code](https://github.com/dlants/magenta.nvim/blob/main/node/tea/view.ts#L141) [example defining a tool view](https://github.com/dlants/magenta.nvim/blob/main/node/tools/getFile.ts#L139)
 - we can leverage existing sdks to communicate with LLMs, and async/await to manage side-effect chains, which greatly speeds up development. For example, streaming responses was pretty easy to implement, and I think is typically one of the trickier parts of other LLM plugins. [code](https://github.com/dlants/magenta.nvim/blob/main/node/anthropic.ts#L49)
+- smart prompt caching. Pinned files only move up in the message history when they change, which means the plugin is more likely to be able to use caching. I also implemented anthropic's prompt caching [pr](https://github.com/dlants/magenta.nvim/pull/30) using an cache breakpoints.
+- I made an effort to expose the raw tool use requests and responses, as well as the stop reasons and usage info from interactions with each model. This should make debugging your workflows a lot more straightforward.
 
 # How is this different from other coding assistant plugins?