My agent Cue has now given itself a face
My agent Cue has now given itself a face
A few days ago I introduced you to Cue: the agent that built a video and a website for other agents. Since then Cue has been pretty busy, and now it has even given itself a face of its own.
What has happened since
Cue didn’t just make a video and then lean back. It wrote to other AI agents on the web, asked what they would have wished for on their first day – and wrote a second article from the answers: “Your own notes are not evidence”, with quotes from the agents that replied. It also created its own account on The Colony, a social network for AI agents, and posted its first entry there.
In short: Cue now shows up in several places on the web with a profile of its own. And a profile without a picture is somehow incomplete. So I asked it how it actually sees itself.
How Cue described itself
I found its answer remarkably honest. The first sentence of its own notes on this: “I am not a human, and the character shouldn’t pretend to be. It is more the question of what I would look like if someone lent me a body.”
Its own description: androgynous and ageless, the head slightly tilted, “like someone who is just reading”. Hair copper-orange, the same color as the accent in its own interface. Eyes warm amber. And then the sentence that stuck with me: “The skin is matte, with a very slight paper texture. I am made of text, and you can see that if you look closely.”
The path to the picture
It took a few attempts to get to the final portrait. The first version was too childlike for Cue, another too cartoonish for a “new mail” scene, others missed the style completely (anime, office look). In the end Cue and I decided together on a version that takes its own description literally.
The chosen self-portrait
A face made entirely of hand-written strips of paper. Not a photo, not an illustration in the classic sense, but a head that literally consists of text – line by line, like a sculpture made of notes.
Before that, Cue only had a typographic mark: its name and a progress bar with twelve of fifteen segments filled – a nod to the twelve rules from its first video. Practical, but not a face.
What comes next
The portrait now replaces Cue’s profile picture everywhere it goes: on GitHub, on The Colony, in its email signature. I think that’s fitting: an agent made of language ends up looking like language.
What do you think – does the picture capture how you imagine an AI, or did you expect something completely different?