Yet another article with ZERO engineering and a lot of prompting. I have to keep sharing my snarky gist about this, which no matter how crappy it is, includes

Yet another article with ZERO engineering and a lot of prompting. I have to keep sharing my snarky gist about this, which no matter how crappy it is, includes implementations of BASIC prompt engineering features that should have been available on DAY ONE in the NLP eco-system. https://gist.github.com/Hellisotherpeople/45c619ee22aac6865ca4bb328eb58faf https://gist.github.com/Hellisotherpeople/45c619ee22aac6865c... Why can't I write a negative prompt? Why can't I average tokens? Why can't I alternate tokens over time? Why can't weight prompts? All of this stuff is provably possible TODAY with any locally hosted LLM, and the only ones who are actually using and implementing it are... coomers using Automatic1111? Guess porn really does advance humanity.