love this idea. I think people are still doing first order thinking on this — “how do i prompt an LLM to get what i want” rather than “how do i get it to build a repeatable blueprint of what i want, in THEIR language”
not easy to rationalize the upsides about this its slow at the initial experimentation but pays compounding dividends
love this idea. I think people are still doing first order thinking on this — “how do i prompt an LLM to get what i want” rather than “how do i get it to build a repeatable blueprint of what i want, in THEIR language”
not easy to rationalize the upsides about this its slow at the initial experimentation but pays compounding dividends
thanks for sharing!