Jev's particular architecture is such that context can be filled faster than with LLMs.