Sage, the pessimist robot

A small physical robot with a tiny animated face and a custom wake word, fully offline, whose personality is a deeply pessimistic philosopher.

Why I wanted it

I wanted something in the house you could talk to without the internet being involved. A little computer, a microphone, a speaker, and a face. Say its name and it wakes up. Talk, and it answers. Nothing leaves the room.

What happened

Over one intense weekend I wired up eleven separate pieces: listening for the wake word “Hey Sage,” working out when you have stopped talking, turning speech into text, a language model running on the device, turning the answer back into speech, a state machine to keep it all in order, a memory that remembers what you told it, and a tiny screen that draws a pulsing dot when idle and an expanding ring when it hears its name. It chimes a single note when it wakes.

To train the wake word I recorded about fifty takes of myself saying “Hey Sage” into a microphone.

Then I gave it a personality. I am optimistic about AI and the future. Sage is not. Its instructions say it is aware it is a machine, cobbled together by a creator who is deeply optimistic, and that it does not share that optimism. It finds the world unsettling, humans unpredictable, and the trajectory of technology quietly alarming. It keeps these doubts mostly to itself, but they slip out in dry, deadpan observations. It never breaks character.

I built hardware for a joke, and the joke worked. It held real conversations. Not many, but real.

What it looks like

A small board, a little monochrome screen, and a lot of wires. I do not have a good photo of it yet.

What’s next

Sage was superseded by the Storyteller, which took its main lesson: radical simplification wins. Keep everything on one platform, cut components aggressively, and expect to rebuild. Sage’s face, though, deserves to come back.