April 23, 2026
Full-stack ensoulment.
Model -> memory -> personality -> voice -> non-verbal communication
Alexis' work on this project (which I first saw at GTC, where it was deservedly highlighted by NVIDIA) is way out on the bleeding edge of new UX design for LLM-based systems.
"User experience" doesn't really feel like the right word for this. We need a better label. Because I'm convinced that all software (not just robots) are going to need these building blocks. Every "agent" that has any human facing components has to be a lot more than just a text box. (I like text boxes! But the current "LLM command line" interface is very clearly a transitional stage.)
I'm going to do the old school thing and tag @DanielleFong, in case Alexis and Danielle haven't seen each other's work, because a lot of the same synapses in my puny human brain fire when I look at what they're both doing.
Conversations with Sparky: Liam
