Feeding an output buffer back as its own input makes tiny changes compound into emergent patterns

Video feedback is a loop: camera to screen to camera to screen, or in code, an output buffer used as its own input. Each cycle adds a tiny modification (a camera can never hold perfectly still; blending is never exact), and those modifications compound over many frames into emergent complexity such as spirals and endless tunnels that neither camera nor screen ‘contains’ on its own. In Hydra this is done with src(o0).blend(src(o0), amount).out(o0); the blend amount controls how much of the previous frame bleeds into the current one. Because the effect is iterated, a small parameter change has a disproportionately large visual result.

Examples

src(o0).blend(src(o0), 0.999).out(o0) keeps almost all of the prior frame, giving ghostly smearing; near 0 it shows mostly the live camera. Adding a slight scale each frame turns the feedback into a zoom or spiral.

Assessment

Explain the feedback mechanism in Hydra via the blend parameter. Why does 0.999 vs 0.998 produce a large difference in the result?