A lot what has been heaped upon us is accidental complexity in Brook‘s silber bullet sense. LLM‘s cut straight through it, collapsing the silos of knowledge (and eco systems) build around languages and frameworks.
I view them as lossy compressors. Kind of like a JPEG for algorithms. And while our optical nerve stack tends to be pretty good at glossing over and correcting for erroneous pixels, it has been my experience that computers do not possess that kind of discernment. Thy do exactly as they're told. And a "pixel off" here and there in the execution of boolean logic can quietly not matter, or it can unravel pretty epically.