Forwarded from † kkyk
This media is not supported in your browser
VIEW IN TELEGRAM
They like it when you look like this
😁4
Everyone "Rigged it"... nah well good game Argentina. and for y'all getting yo points straight! ok spain don't fuck this up na.
🔥1
This media is not supported in your browser
VIEW IN TELEGRAM
was already working on Argentine.
🔥2😁2❤1
Dagmawi Babi
I talk to a lot of people building AI products, and are actually good AI/ML engineers, I also talk to some really cool devs and most of them have some form of personal AI. For me I use AI to analyze my personal content and generate markdown reports, for some…
i asked my "personal AI", Aide, how much it would rate itself and environment based off of babi's post... The honest take is higher than i thought.
Aide has been upgraded several times, with the help of Aide its self sometimes, to make it as dynamic as possible. while this response sums it up there are alot of things it can do with the tools it have... it can literally replace a subscription tracker with one sentence right now.
it has also evolved from Routing (where a Larger LLM routes decisions to another cheaper one for tool calling) to a single model ReAct loop (Reasoning and Action) because of the lossy compression of context that happens between models when efficiency is at hand... it still uses 2 models but the smaller model is only called upon 30min after a conversation between the user and the deep model takes place... then it analyzes the convo and categorizes and saves anything else to its file and build profile of the user
i was thinking of adding voice as in "ambient access" lately... we will see how it develops tho.
Aide has been upgraded several times, with the help of Aide its self sometimes, to make it as dynamic as possible. while this response sums it up there are alot of things it can do with the tools it have... it can literally replace a subscription tracker with one sentence right now.
it has also evolved from Routing (where a Larger LLM routes decisions to another cheaper one for tool calling) to a single model ReAct loop (Reasoning and Action) because of the lossy compression of context that happens between models when efficiency is at hand... it still uses 2 models but the smaller model is only called upon 30min after a conversation between the user and the deep model takes place... then it analyzes the convo and categorizes and saves anything else to its file and build profile of the user
i was thinking of adding voice as in "ambient access" lately... we will see how it develops tho.
🔥2