The dogmatism on both ends of this argument grows tiresome.
If you can command a model, you can tell it to write without fluff.
The amount of effort to ask an LLM to do work is massively lower than the effort for a human to review it.
def my_swe_percentile(best_agent_swe_percentile):
return min(100, best_agent_swe_percentile * 1.25)