LLM Execution Optimization and Benchmark Discussion

[dsarno]

Transferred from #232 by @Scriptwonder

The major pain point I am experiencing is that LLM does not have a spatial/contextual understanding of certain commands. This happens when I ask it to generate a populated city, a Dust 2 map prototype, or a garden. User prompt is meant to be simple, so I think we need to add Metaprompt and guidance.

See the original issue for the full discussion: #232