joey.hersh

systems / rayneo-hud

RayNeo iO HUD

Smart glasses as a heads-up display for my agents, messages, media, and home

The phone is a radio. The glasses are the screen. The agent is the app.

Period
November 2026
My role
Creator. Designed the screens and control scheme; built the Android companion, the Bluetooth transport, the gateway, and the companion apps; brought it up on custom firmware.
State
My daily driver, running on Turbo-IO custom firmware. Open source.
  • 16Screensagent, messages, coding threads, media, calendar, health, home, and more
  • 9Services behind one gatewayHermes, T3, Beeper, Spotify, Calendar, Oura, Home Assistant, Audiobookshelf, transcription
  • 75Screen states with pixel referencesKotlin renderer matches the HTML simulator to within 0.1%
  • 50Joy avatar animationsthree frames each, packed to one bit
  • Kotlin
  • Android
  • Bluetooth LE
  • RFCOMM
  • Protobuf
  • Node.js
  • Docker
  • Cloudflare Tunnel
  • Hermes
  • Home Assistant

Source on GitHub ↗

isometric smart glasses with a floating heads-up panel in one lens, linked wirelessly to a small pocket controller
The glasses are the screen; an 8BitDo Micro in a pocket is the controller.
rayneo.hud: architecture flowgolden imagesHTTPS8BitDo Micropocket controller · push-to-talkpocket controller · push-…Android companioninput · screens · rendererinput · screens · rendererBluetooth transportBLE login · RFCOMM · protobufBLE login · RFCOMM · prot…RayNeo iOTurbo-IO firmware · 540×220Turbo-IO firmware · 540×2…HTML simulatorpixel-parity referencepixel-parity referencejoy-api gatewayone container · Cloudflare tunnelone container · Cloudflar…Joy on Hermesagent · voice · memoryagent · voice · memoryServicesBeeper · T3 · Spotify · HA · OuraBeeper · T3 · Spotify · H…

Diagram nodes, in flow order: 8BitDo Micro (pocket controller · push-to-talk); Android companion (input · screens · renderer); Bluetooth transport (BLE login · RFCOMM · protobuf); RayNeo iO (Turbo-IO firmware · 540×220); HTML simulator (pixel-parity reference); joy-api gateway (one container · Cloudflare tunnel); Joy on Hermes (agent · voice · memory); Services (Beeper · T3 · Spotify · HA · Oura).

deterministic code model judgment human durable state tool / API

Why

Most of what I do on a computer comes down to four things: talking to my agent, managing coding-agent threads, answering messages, and playing media. Every one of those already runs through an API I control. So I bought a pair of RayNeo iO glasses and an 8BitDo Micro, and replaced most of my phone with a screen in front of my eye and a controller in my pocket.

How it works

The phone app owns everything: controller input, a screen stack, a renderer, and the transport. Every screen draws twice, once as a 540 by 220 image for custom firmware and once as a 26 by 7 character grid for stock firmware, so the same app drives any pair of glasses.

The transport speaks the glasses' own protocol: a Bluetooth LE login, a classic RFCOMM serial channel, and protobuf messages across twelve channels. I built it in Kotlin on top of the community's protocol research and hardened it against what the hardware actually does in the field, like GATT operations that come back busy and need a retry, and a caption page that accepts a microphone start but never streams. On the Turbo-IO R1 build the assistant recorder streams microphone audio within 300 milliseconds, and full raster screens go out one to one.

Screens are designed first in a single-file HTML simulator. The Kotlin renderer redraws the same 75 screen states, and tests compare the two pixel by pixel. The worst case is under a tenth of a percent.

The gateway

The glasses never talk to a service directly. One Node container, joy-api, sits behind a Cloudflare tunnel on my home server and connects to Hermes, T3 coding agents, Beeper, Spotify, Google Calendar, Oura, Home Assistant, Audiobookshelf, and transcription. A new integration is a module on the gateway and a screen in the simulator.

Companion apps

The same platform carries a bike-navigation app with voice place search and turn-by-turn directions, and a skate companion with ride tracking, Strava export, and board-light control. Each is its own library with its own tests that plugs into the phone app.

six RayNeo iO HUD screens at native 540 by 220, a Codex thread, Joy, a message thread, an audiobook, the morning report, and Home Assistant scenes
six RayNeo iO HUD screens at native 540 by 220, a Codex thread, Joy, a message thread, an audiobook, the morning report, and Home Assistant scenes