You mean all these testaments are bs? As an infrequent user of LLM assisted work these stories never really tallied with my experience.
For example I could never throw a bunch of spec/doc at an agent and have it return something useable 30 minutes later. Yeah the code compiles but they don’t work.
Either they're BS, or the people making these statements are self-incriminating to a terrible degree, either they don't care about their work or are outputting a very low level of quality and being amazed at how "great" and how much better AI output is than their own
For example I could never throw a bunch of spec/doc at an agent and have it return something useable 30 minutes later. Yeah the code compiles but they don’t work.