Hans

The best results I've found come from "crossing the streams," so to speak. As others mentioned, to take a spec from one GPT and apply it with another. I've found that breaking things into fine chunks, ideally "hopped" between GPTs, provides dramatically decreased errors, and increased end-to-end function.

While I agree things are too complicated, it's a symptom of the mess that the web stack itself has become. The web is for transmitting information and the layering of junk comes from it being manipulated to do something it wasn't meant for - there shouldn't be security concerns when a site also has to pay for SSL - why is my bank in the same category as some random site? It's been a false...