diff --git a/src/pages/blog/roadmap.md b/src/pages/blog/roadmap.md
new file mode 100644
index 0000000..2708b4a
--- /dev/null
+++ b/src/pages/blog/roadmap.md
@@ -0,0 +1,87 @@
+# Roadmap
+If you want to add or prioritize an item, hit me up on Discord.
+
+If something is prefixed with (HANG) then it might start closed source; I'm not sure.
+Everything possible with stock WebRTC *will* be open-source.
+But I do want to make an experimental product with niche features that don't make sense to open source immediately.
+Maybe when my startup implodes; hit me up if you're interested.
+
+## Phase -1: Already Implemented
+This project has been meandering so I don't blame you for being unaware of what has already been completed:
+
+- moq-relay
+  - Server that fans-out publishers to subscribers.
+  - Supports clustering, fetching content from other relays around the world.
+  - Media agnostic. Doesn't care about the content of each track.
+- moq-karp
+  - Media "container" on top of moq-transfork.
+  - Mostly consists of a JSON catalog.
+  - CLI to convert and publish fMP4 streams.
+- moq-web
+  - Web playback and publishing.
+  - WebComponents and Typescript API.
+  - Core is in Rust (WASM+Worker). 
+  - Video only: h264, h265, av1, vp8, vp9.
+  
+
+## Phase 0: Foundation 
+Before diving headfirst into the product, I want to "finalize" the base so it can be built upon.
+But I don't want to revert to arguing semantics at IETF.
+
+- Refocus the base drafts:
+  - Rename `moq-transfork` as `moq-lite`. The focus is simplicity and live only.
+  - Publish `moq-karp` as `moq-hang`. The focus is live conferencing only.
+- Implement `gstreamer` plugins:
+  - Implement a `src` for receiving media over the network.
+  - Improve the `sink` for sending media over the network (no more fMP4).
+- Resurrect `moq-js` for web.
+  - It was a fun experiment, but the web isn't designed for Rust.
+  - Simpler, more performant, and less complicated web client.
+  - Already has (poor) audio support.
+  - Rename `kixelated/moq-rs` to `kixelated/moq` with a `native` and `web` folder.
+
+## Phase 1: Core Product 
+The basics required to have a voice chat with someone online.
+
+- Improve web audio support:
+  - Investigate getting rid of `SharedArrayBuffer`.
+  - With the ability to mute/deafen/volume.
+- Implement a room UI:
+  - Create/join room button.
+  - Grid of connected users.
+  - Configurable channel display name.
+  - Configurable (saved) user display names.
+- Implement a simple chat protocol:
+  - No history
+ 
+## Phase 2: Branding
+Improve the quic.video site so people can actually try this stuff.
+
+- Improve the design so it looks less shit.
+- Add a meeting demo.
+  - List public rooms.
+  - Allow creating a new room.
+- Slowly release a cache of hoarded blog posts.
+- Start attending IETF meetings again.
+
+## Phase 3: Hang
+Create hang.live as a true demo of what MoQ can do.
+
+- (HANG) Replace the grid with a dynamic canvas.
+  - No spoilers
+- (HANG) Add notifications.
+- (HANG) Improve screen sharing.
+- (HANG) Mess around with shaders.
+
+## Phase 4: Polish
+
+## Phase N - Future?
+
+- Implement an OBS plugin:
+  - Should be easy, but low priority.
+- Improve the ffmpeg integration:
+  - Investigate not using pipes and an fMP4 intermediate.
+  - Add the ability to receive live media.
+- Clean up the async internals:
+  - Reimplement the core as a synchronous state machine.
+  - Support Quiche as a QUIC backend.