AI is code, and can't be prompted into being smarter

Article excerpt

A technical argument challenges a widespread belief about AI capabilities: that better prompting techniques can unlock hidden intelligence in language models. The piece contends that AI systems are fundamentally code, software with fixed parameters determined during training, and therefore cannot be coaxed into performing beyond their trained capacity through clever instruction alone. This reframes the ongoing debate about prompt engineering from a path to AI improvement into something closer to optimizing output from a static system. The argument suggests that meaningful advances require retraining or architectural changes, not conversational finesse.