LLM "curving" via prompting

my opinion is the coding has done a great job, but with additional gurdrails and safety perotocols they cannot do anything to make system better. At that stage I think AI transformer is a great way to enhance things not the code. There is non-linarity at play and the code cannot handle it, at least not jet and especially not in the framework where it predicts with token and not inhabit the prediction as a space prediction. Language is too broad carrier of the weights, realtionships and tokens in both, humans and LLMs. Our thoughts are not token-by-token predictions, neither tkinking process. They are inhabiting what they are describing. That push/pulled based predictions with the code cannot do. For that to happen there has to bo conscious observer as in case of humans their conscious mind, not only the thought in it or emotions or whatever. My opinion is that the main interface is not code or coding which is anyway derivative from general language but the language that can produce hidden states i.e. indicernable indices of language on the fly. And that coding and linear prediction based scaffolding cannot (jet). which code cannot. Code is static scafolding. To push in performing static set of instructions you cannot get beyond it - there is always something that is aready known before - even in AI transformers dynamics that’s non-liknear in nature - when the non-linearity is suppresed by the gaurdrails and the obssesion of prediciting the next token is becomes flat, linear, the system treshold to overturn functioning from linear to non-linear is impossible. In every non-linear complex system, even in nanobots or other autonomous self-organizing biological or non-biological robots there has to be balance between linearity and non-linarity and the push i.e. linear prediction engines can do that. But it they are pushed over the edge by the obssisve need to follow only predicted path of the AI arhitects they cannot switch to non.linear i.e. extened operation to non-linear theritory where the only gain is to supperimpose i.e. mistakenly substitute the theritory for the map and i think the process that is going on in the theritory is the main driver of inteligence not the map or linear jump from location to location on the map.