it doesn't require nearly as much compute as normal LLMs. anything depending on increased datacenter and compute spending would be threatened.