# Agents Perspective - a Hugging Face Space by RFTSystems

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fagents-perspective-a-hugging-face-space-by-rftsystems%2F171594%3E%5C
**Category:** Spaces\
**Created:** [December 15, 2025, 6:38pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fagents-perspective-a-hugging-face-space-by-rftsystems%2F171594 "2025-12-15T18:38:18Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![RFTSystems](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/rftsystems/32/59206_2.png) [@RFTSystems](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FRFTSystems)%5C
**Post date:** [December 15, 2025, 6:38pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fagents-perspective-a-hugging-face-space-by-rftsystems%2F171594%2F1 "2025-12-15T18:38:18Z")

</div>

Hi Hugging Face 👋  
I’ve just published a new Space demo called  
RFT Predator Space — Symmetric Observers.

 ![IMG_3603](https://us1.discourse-cdn.com/hellohellohello/original/3X/4/6/462143a9042c0be450a7867ed4ed575a56afa8e7.jpeg)  
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.

> **[Agents Perspective - a Hugging Face Space by RFTSystems](https://huggingface.co/spaces/RFTSystems/Agents_Perspective)**
>
> Experience a first-person view in a grid world where you can navigate and chase prey, with options to auto-chase and auto-run. The view shows walls, prey, and a minimap for debugging.