A design proposal for a monitor-top robot that keeps you in peak performance with helpful reminders.
It notices when you sit down, then nudges you to take a break, drink some water, grab a snack or sit up straight. It uses a puff of air and a friendly voice, so you never have to look away from the game.
An AI render of the CAD model. This is a concept, designed down to the circuit board but not yet built.
Headphones on and eyes locked on the screen. Three hours vanish, the water bottle is still full and their back is a question mark. A chibi robot that looks like it belongs next to their RGB keyboard is something they'll actually leave on the monitor.
They've asked nicely through the door four times. Gamer Guardian repeats those reminders from inside the kid's own setup, in a voice that isn't Mum's or Dad's, and it doesn't need a camera in the bedroom to do it.
The moment: 90 minutes into a ranked match, a cool breeze hits your face and a small voice says "Grab some water. You've got this."
Pick a reminder. The robot's eyes light up in that reminder's colour, it speaks one short line and then the breeze kicks in. Turn voice on to hear it in your browser's own speech voice. The real robot plays recorded phrases.
Each reminder has its own eye colour, so a glance from the corner of your eye is enough.
On the real robot, saying "done", "snooze" or "I drank water" is heard by a wake word and fixed-command model that runs on the chip. No audio leaves the device.
This is an example schedule, not a tested one. The timers pause whenever the PIR sees no one at the desk, and the robot goes back to sleep.
A 30 mm blower behind the mouth grille puffs a cool breeze at you for a few seconds. You feel it through a headset, mid-fight, without it covering any game audio. It runs at a low PWM duty so it whispers rather than whines.
A 28 mm speaker on the side plays one stored phrase: "Time for a break! Stretch your body." It's loud enough to get past headphones at speech level, and short enough not to break your concentration.
Two RGB LEDs behind the smoked visor stay dark the rest of the time. When a reminder fires they glow in its colour, so peripheral vision tells you which one it was. They never blink on just to look pretty.
Privacy is part of the design. Presence comes from a passive infrared sensor, the same kind used in a hallway light, so it knows someone is there but not who. The microphone listens for a wake word and a short list of commands, all on the ESP32-S3 itself. There's no camera, no cloud account and no recordings.
This is the first-pass CAD model, with every dimension taken from real, cheap parts. Drag to spin it. It has a face cap and a rear shell, a smoked PMMA visor, neon side rings, TPU feet and a hinged clip that hooks down the back of the monitor.
One 54 × 31 mm two-layer board stands upright behind the face. The front of the board is what sees out through the visor: the PIR and the two eye LEDs. The back carries the ESP32-S3, the audio and the power parts. The blower, the speaker and the USB-C socket connect with short JST-SH leads.
The speaker sits on the right wall, out of this plane. Every gap is a first-pass guess that needs checking on a print.
We picked the S3 because it's the ESP32 that can run Espressif's ESP-SR speech stack (a wake word plus about 200 fixed commands) on the chip itself. It also has two I²S ports, one for the mic and one for the amp, and native USB, so it flashes over the same cable that powers it.
A tiny 3 V PIR draws about 15 µA. Its dome pokes through the visor, and it gets its own RC-filtered supply so fan and Wi-Fi noise don't cause false triggers.
A digital I²S MEMS mic, ported through a 1 mm hole in the board to a pinhole in the face.
A class-D I²S amp on 5 V drives a 28 mm, 8 Ω, 1 W speaker. It needs no codec and no filter.
A low-side MOSFET switches the blower at 25 kHz PWM, which is too high to hear. A flyback diode protects it, and a pull-down keeps the fan off at boot.
Two 2 mm RGB LEDs on one data line. One diode drops their supply to about 4.3 V, so the 3.3 V data signal reads reliably and no level shifter is needed.
USB-C 5 V in, with a 600 mA LDO for the 3.3 V rail. There's no battery, so nothing ever needs charging.
The green bar is typical draw and the pale bar is peak. The orange line marks the 500 mA a USB 2.0 port promises.
If everything peaked at once it would draw about 870 mA, so the firmware takes turns: speech first, then the breeze, with Wi-Fi off. On a USB-C charger, the 5.1 kΩ CC resistors unlock 1.5 A anyway.
| Ref | Qty | Part | What it does | MPN |
|---|
This is a speedrun first pass. Here's what we'd want to answer before anyone orders a board or cuts a mould.
Everything behind Gamer Guardian is open: the 3D design, the KiCad boards and the notes. Download it, change it, make it, even sell it. All we ask is a credit and a link back.
Open the project on GitHubThe build tools
Enclosure, ready to edit or print
KiCad project, schematic and BOM. Saved with KiCad 10: open it in 10 or newer.
Paste this into Claude Code, Codex, Cursor or any agent that can run commands. It fetches just this project and the tools, reads the notes in the right order, and asks you before it changes anything.
Hardware under CERN-OHL-P v2, words and images under CC BY 4.0, code under MIT.
A short diary of every episode: the roll, the turns the idea took, and the board at the end. One email per episode.
