Baseline gravity
Stars orbit in a softened analytic disk+bulge potential integrated with a fixed semi-implicit step. This is intentionally stable and deterministic rather than a direct N-body solver.
Gas hybrid
A coarse rotating cellular gas disk couples to capped SPH-lite clouds. Dense gas can form tracer stars; scheduled feedback disperses nearby gas and later cooling allows the cycle to resume.
Experimental fuzzy halo
A low-resolution complex field evolves with a bounded finite-difference Schrödinger-like step, then a smoothed density field contributes a gentle acceleration. It is a visualization-grade approximation, not a research solver.
Gravity paint
Drag on the galaxy to add a smoothed external potential. The painted field is clamped, diffused, and differentiated; its acceleration contribution is hard-limited to prevent singularities.
Structure tracker
Angular/radial overdensity peaks are matched over analysis frames into stable IDs. Persistence is informative only until mastery unlocks exactly one payoff: a short predictive structure projection overlay.
Synthetic telescope
The telescope view changes only observation: blur, extinction, detector noise, and LOS velocity coloration. Hidden State withholds privileged overlays and asks you to infer a real property of the live deterministic state.
Failure modes
The gas, halo and structure models are intentionally coarse. Strong painting, very low quality, or extreme speeds can create approximation error. Diagnostics expose invalid-state recovery and halo residuals rather than hiding them.
Try this first
Start Full Hybrid at seed 8841. Watch an arm stabilize, paint a gentle attractor just outside it, switch to Telescope, then Hidden State. Increase Bloom for a cinematic plate or reduce it for analysis. Finally run the A/B benchmark and compare structural persistence and drift.