Hi Hugging Face 
I’ve just published a new Space demo called
RFT Predator Space — Symmetric Observers.
The goal is to make one idea immediately visible and testable: two independent observers can co-exist in the same frame, with symmetric viewpoints and policies, without relying on black-box models. What it is: a deterministic predator/prey gridworld rendered in pseudo-3D first-person. You can toggle control between Predator and Prey (camera + input swap), so you literally see the environment from either observer’s perspective. Visibility is honest: the other agent only appears when line-of-sight + field-of-view allow it, and is correctly occluded by walls (no “see through walls” shortcuts). Modes included: Manual play, AutoRun+AutoChase, and a Hybrid mode (AutoRun ON, AutoChase OFF) where the predator wanders autonomously while the prey flees — showing two independent observers running their own policies in the same world. Quality/safety note: I removed UI flicker by making the progression panel event-driven (no flashing updates per tick) to avoid photosensitive risk. There’s also an optional subtle “coherence/disturbance” overlay (off by default) that reacts to sharp turns / bumps. Features: deterministic seeds for reproducibility, unlockable maps as your catches increase, and save/load (slot saves + export/import JSON). If anyone wants to stress-test it, I’d love feedback on: (1) the perception rules (LOS/FOV/occlusion), (2) whether Hybrid mode reads as “two observers” in practice, and (3) any map ideas you’d want as unlocks.