Earlier quoted context omitted.
Please explain this "adapting". The LLM sellers' holygrail is * you just need to crank our model bro *. With this mindset, there is no adaptation needed, save a couple of weeks of getting up to speed with nuances of using the models, the / commands and integration of their APIs into your workflow.
You gotta ignore what the LLM sellers are saying and listen to the people who don't have a financial incentive to sell a particular message.
It's asking engineers to "adapt" to being efficient in filling out the last 10% then 1% then 0.1% then 0.01%.