TheFan83

open.spotify.com/intl-de/alb… If you like poker, if you like rap, check out my just released album Poker Truths

"Live now", the best life advice of my life and i keep forgetting it. My body insta relaxes and happiness up when i do remember it.

there is a big difference in thinking you are talking to a less then average common senes human being or one above the average..

gpt 6.1 sol : 6bb/100 loss vs gto over 100 hands it played to use up its 5h limit. claude sonnet 5.5: 35bb/100 loss vs gto over 200 hand it played within that limit (and faster) You dont get to 6bb/100 with buttonclicking, really impressive imo. (4 sizings bu vs bb srp 100bb)

Back to using opus 5.1 ... i have done 5 revisions of a task in the time i watied 3 5 hour windows to have then somemthign i dont want because i did not dare to comment at all to not burn more limit with astra...

I dont care what cool 3d things astra can do, i still consider it dumb so far.. as long as i have to scream in my mind "why would you ever??"..

can @openai just let 10.000 super agents lose and create GTA 7? or whatever @sama likes to play, seems like there is interest in this and with enough time and agents.. would be interesting.

One thing that has not changed since the orginal chat gpt release.. models feel incredibly dumb around the time the next model comes around and then you are like "woahhhh, this is agi now"

if astra would be free, i would have it do SO much its insane... but it is not. So here i am sitting waiting for my 5h window.

How is GPT-6-Astra supposed to fix all the mess the older models did over the last months if it eats 5 hour windows for breakfast?

aigamesforyou.com/games/giza… 2.3 million stones, GPT-6-Astra did them in 15minutes

caveat

in terms of ai progress we have seen nothing yet, claude mythos / fable is great, BUT it is at the border or slightly below of experts at niche tasks overall imo, so NOT speeding up progress all that much. Once we cross the border where its just more intelligent than humans..