Post #896962

The thing that really gets me about how all these LLM prompts are written, especially Claude, is that they're *IN ENGLISH*. You would think that they might be expressed more precisely and effectively as something in a vector space, native to the model. Nope. Just in English, instructing a mindless, unintelligent algorithm to *hopefully* emit a different token sequence. None of this shit is coding. It's not anything. This is the upside-down hellscape inverse of coding.