The robot woke up this week. NEURA has lived on my desktop as software, but this is the first time the whole brain ran on the actual robot. I got it running on a Raspberry Pi tucked behind a small round screen, set to boot straight into a face I can wake up.
The round screen was the first real fight. Everything I had built assumed a rectangle, so the eyes and the UI kept getting clipped by the bezel, corners sliced off, text running into the curve. I built a round-screen mode that lays the face out to fit inside the circle instead of fighting it.
Then I went down to the hardware and wrote the driver for the motor controller. I verified it all the way down, not just in code but actually on the bench. There is a specific satisfaction in watching software you wrote turn into motion you can feel.
The other big piece was survival. During testing the robot gets its power yanked constantly, and every yank used to leave it dead or stuck on a blank screen. So I built a resilience kit: if the face crashes or the whole thing loses power, it relaunches itself and comes back to a clean, wakeable face on its own.
That matters more than it sounds. A companion that only works when I am there to restart it is not a companion, it is a demo. The point of NEURA is that it just sits there, ready. Getting it to always return to a clean face by itself is what makes it feel like a real device.
It is still rough. But there is a round little face on my desk that boots up, waits, and wakes when I talk to it, and that is a different feeling than any milestone before it.