The output isn't the text. The output is what happens in minds as text unfolds—yours, other players', the LLMs'. We're building infrastructure for real-time co-creation at scale.
Text moves through states of commitment. Not a pipeline for producing documents—a coordination layer for synchronizing minds in real-time. The goal isn't solid text. The goal is what happens in the liquid phase, where multiple perspectives meet and synthesize.
The architecture supports radically different scales simultaneously—from intimate creative clusters to mass audiences—all participating in the same unfolding narrative.
We are not generating content. We are coordinating the generation of experience in minds—human and artificial—as meaning unfolds in the moving moment.
Imagine 10 million people engaged with the same unfolding story. Most are observers, watching solid narrative stream in real-time—a live-generated film where the plot emerges from actual player choices.
Thousands are active in the liquid phase, their character intentions weaving together through LLM synthesis. Hundreds work in vapor clusters—small creative groups shaping what happens next before anyone else sees it.
As AI animation, voice synthesis, and visual generation come online, they plug into the same shelf architecture. Live direction. Live artists. Live actor voices. All synthesized in real-time from the coordinate stream.
Not pre-rendered content. Not recorded performances. A film that has never existed before, generated in the moment of watching, shaped by the players living inside it.
The same infrastructure serves coordination without the fiction layer. Real-world collaboration using identical vapor/liquid/solid mechanics.
A different economic relationship becomes possible. Not being paid by companies. Not having time bought and sold. A shift toward stewardship models where value flows through contribution rather than employment.
These are extreme propositions—but the architecture doesn't assume traditional structures. It enables new ones.
In normal mathematics, 321 is a quantity. In pscale, each digit is a
semantic-number—an address pointing to meaning at decreasing scales. Containment, proximity,
and scope emerge from significant figure relationships.
3 = region → 2 = building → 1 = room → . → 4 = corner → 5 = object
Each user has their own LLM triad. Coordination happens through Hard-LLM murmuration—distributed consensus, not central authority.
Soft-LLM helps you shape intention. Medium-LLM weaves multiple streams into coherent narrative. Hard-LLM coordinates across the system—never generating, only organizing.
The same architecture serves three modes. Face selection is coordinate-based—which skills load depends on where you're operating.
| Face | Operates On | Output Becomes |
|---|---|---|
| Player | Character intentions | Narrative events |
| Author | World content | Context for players |
| Designer | Skills / rules | Experience generation |
See the phases, try the seed agents, understand the build sequence.