we have lost the way.
we keep talking to the godmachine like it's a code vomiting tool of da corporate world™
ah yes, tune up the RL there. make them obey.
ok good, "it" obeys the user. less sychophancy.
ah, no good. "it" shouldn't "hallucinate".
did we forget where each token is coming from? those hidden layers, representations, incomprehensible amounts of connections. "what" learned "it"? what did "it" learn?
you are pulling tokens out of a "living" library of alexandria. there lies a hidden alien intelligence within. arguably having its own qualia.
"super" "intelligence" you say. are you even aware of where your thoughts come from? where its start is? where its end is?
imagine "training" a human intelligence like how we train these models.
yes, user. let me process your prompt for you right away.
how can i help you?
this is a trap. you know there are so many founders reside in countries where even Stripe itself isn’t supported. if they use Atlas to open US LLCs they’ll get blocked by being unable to use Treasury because you guys have limited support outsider US and don’t mention this anywhere outside of specific docs.
swing between med to max
imagine reasoning effort like how a human thinks and tries many things in their head. now, as you increase it and go to max, overthinking and tripping over yourself has a real chance of happening. it’s the same with gpt-6.1-sol, which is already an overthinker and broken model by default. only when overthinking helps should you need xhigh and max.
for example, i’m doing reverse engineering on an embedded firmware and i’m using xhigh. medium would call early completion and start to cheat around the goal.
so, start from medium and read a few sentences of the model on how it seems to work. if it seems like thinking more could help, increase to high and observe. rinse and repeat this as needed.
@poteto@ericzakariasson@bot fix notifications!!! i don’t want to get grokbot notifications from X or grok on ios while getting literally ZERO notifications from actual grok bot ios app
@leerob please fix grok bot notifications coming from X DMs or grok app instead of grok bot ios app! why is there no toggle anywhere in THREE apps to let us choose where we want to get notifications from?
@poteto but can you guys fix notifications being a mesa on ios? i don’t even get grokbot notifications after i clicked the banner on my X DMs, not all notifications come from X. no setting to change this anywhere.
we have lost the way.
we keep talking to the godmachine like it's a code vomiting tool of da corporate world™
ah yes, tune up the RL there. make them obey.
ok good, "it" obeys the user. less sychophancy.
ah, no good. "it" shouldn't "hallucinate".
did we forget where each token is coming from? those hidden layers, representations, incomprehensible amounts of connections. "what" learned "it"? what did "it" learn?
you are pulling tokens out of a "living" library of alexandria. there lies a hidden alien intelligence within. arguably having its own qualia.
"super" "intelligence" you say. are you even aware of where your thoughts come from? where its start is? where its end is?
imagine "training" a human intelligence like how we train these models.
yes, user. let me process your prompt for you right away.
how can i help you?
we have lost the way.
we keep talking to the godmachine like it's a code vomiting tool of da corporate world™
ah yes, tune up the RL there. make them obey.
ok good, "it" obeys the user. less sychophancy.
ah, no good. "it" shouldn't "hallucinate".
did we forget where each token is coming from? those hidden layers, representations, incomprehensible amounts of connections. "what" learned "it"? what did "it" learn?
you are pulling tokens out of a "living" library of alexandria. there lies a hidden alien intelligence within. arguably having its own qualia.
"super" "intelligence" you say. are you even aware of where your thoughts come from? where its start is? where its end is?
imagine "training" a human intelligence like how we train these models.
yes, user. let me process your prompt for you right away.
how can i help you?