Home / Project 02

Project 02 · Hardware · Built in public

BOB is getting a body.

BOB has lived on a server so far — in chats, dashboards and a 3D office. MK-II brings him into the room: a head on the desk with a living eye, a camera, microphones, a voice and a few senses. Same BOB, same memory, same personality. Just physical.

Platform
Raspberry Pi 5 · AI accelerator (Hailo-8)
Brain
Mission Control + Hermes Agent on our server
Status
Stage 1 parts ordered · head design in final review
Next milestone
“BOB speaks” — first spoken conversation
3D model of BOB MK-II: a white, rounded head with a glowing red eye and a camera lens, on a three-axis neck above a dark wooden base3D twin from Mission Control · not yet printed

Why a body

A presence in the room, not another screen.

An assistant you have to open is an app. One that notices you walk in, looks at you when you talk and tells you what the day holds is something else. MK-II is our experiment in what happens when an AI agent gets a place in the physical world.

How it works

Reflexes on the desk, thinking on the server.

  • The head handles everything that has to be instant: the eye, the wake word, where a voice comes from, faces, touch.
  • The thinking happens where BOB already lives: Mission Control and Hermes Agent, with his full long-term memory.
  • The head only opens outgoing, encrypted connections. Nothing in the home network has to be exposed.
  • If the server is unreachable, the eye shows it — and the local reflexes keep working.

The design

A friendly bean with one big eye.

After several studies — from a sharp wedge to a kit-based robot — we landed on a soft, rounded shell. The left “eye” is a round display that can look, blink and show what BOB is doing; the right one is a real camera. The shell is 3D-printed, with a sensor dome on top for the environment sensors, and the head sits on a three-axis neck above a wooden base that houses the computer.

All images are renders of the current design. Photos of the real build will follow.

What's inside

Four ways to sense the world.

Semi-transparent 3D model of BOB MK-II showing the parts inside: the round display eye, camera, speakers and sensors in the head, the neck servos and the computer in the base
Every part at its real position, in the 1:1 digital twin we built in Mission Control.
See

Eye & camera

  • Round 2.8″ display as a living eye
  • 12 MP camera with autofocus
  • Face detection on the AI accelerator
Hear & speak

Voice

  • 4-microphone array with echo cancellation
  • Knows which direction you speak from
  • Speaker with BOB's own voice
Think

Brain

  • Raspberry Pi 5 for the reflexes
  • AI accelerator for vision, on the device
  • Mission Control & Hermes Agent for everything else
Feel

Senses

  • Air quality, CO₂ and fine dust
  • Touch zones and a motion sensor
  • Radar that sees who is in the room

Privacy by design

A camera in the room has to earn trust.

These rules are part of the design from day one, not an afterthought.

  • Camera images never leave the device. Only events like “someone is here” do.
  • Audio is only sent after the wake word — and only that one sentence.
  • The eye and the lights always show when BOB is listening.
  • A voice command or a long press on the back of the head mutes the microphones.
  • Faces are only remembered after explicit introduction, and “forget me” deletes them.

The build

Where we are.

Every step ends with something you can see or hear. The next one only starts when the last one works.

  1. October 2026Design & parts

    Head design in final review, all parts for the first stage ordered, first sensors on the way.

  2. October 2026Digital twin

    A 1:1 3D model of head, neck and base with all 24 parts, plus a motion lab in Mission Control to test every move before the hardware arrives.

  3. NextDisplay test & the eye comes alive

    Check that the round display runs smoothly at 60 fps, then give the eye its states: idle, listening, thinking, speaking.

  4. Milestone ★BOB speaks

    “BOB, what's the weather like?” — and BOB answers with his own voice. From here on, MK-II really is BOB.

  5. ThenBOB sees & senses

    The eye follows faces and voices, BOB recognises the people he has been introduced to, and notices stuffy air or a pat on the head.

  6. Stage 2Movement

    A three-axis neck driven by three servos: the head turns towards whoever is speaking, nods while listening and tilts when it is curious. Every move can already be tested in the digital twin.

  7. Milestone ★Morning briefing

    You walk in in the morning, and BOB greets you with the weather, today's tasks and reminders.

Curious where this goes?

We'll share the build as it happens. Questions, ideas or a project of your own? Write to us.

Get in touch