Was curious to try to outsource hardcoding particle systems to AI - similar to how some codegen graph-based particle systems work.
These produce bytecode / generated code.
In AI case we skip these steps and iterate with agents.

Result is compute-based particles. ~Hour and half of work.

Maybe now is the time to share!

A quick way to reduce bandwidth A LOT when dealing with Deferred Renderer on mobile

Possibly you have non-optimal order of passes when dealing with shadowed lights. This saves 120+MB/frame with 7 lights on iPad

More: www.gamedeveloper.com/game-platfor...