Blog
What changes when an AI has a face
Most AI chat is a rectangle of text that appears a word at a time. It works. It is also the least human way to present something built entirely out of human language.
Give the same model a face — one that blinks, follows your cursor and shifts expression with every reply — and people treat it differently. They say hello. They say thank you. They come back.
A face is state you can read without reading
A text box tells you nothing until the words arrive. A face is already telling you something: it is thinking, it is waiting, it noticed you, it has nothing to do.
That is the practical argument, and it is not sentimental. Expression is a status indicator that needs no label and no spinner. You glance at it and you know where you are.
- Thinking — eyes narrow, attention turns inward
- Curious — wide, alert, head slightly tilted toward you
- Sleepy — half-lidded, slow blink, nobody has spoken in a while
- Surprised — the reply contained something the model did not expect
None of that needs a legend. You have been reading those faces since before you could read words.
Where a face earns its place
Anything someone uses more than once. A tool you open daily is noticeably better company with something looking back. The difference between a dashboard and a companion is mostly eye contact.
Anything with waiting in it. Models take seconds to answer. A spinner makes those seconds feel like latency; a face makes them feel like someone thinking.
Anywhere you want people to start. The hardest part of a chat interface is the empty box. A face that looks up when you arrive is an invitation in a way a blinking cursor is not.
Try one
All ten companions are free to use with no account. They run in your browser on your own Claude or ChatGPT key, so nothing you type passes through anyone else's server — and you can watch an expression change in real time as the answer arrives.