Video yükleniyor...
Video Yüklenemedi
why isn't agent thinking visual?
79,361 görüntüleme • 20 gün önce •via X (Twitter)
33 Yorum
It consumes more tokens :(
might save tokens if they are just a set of available animations
Because no one wants to pay for pictures they'll ignore after 2 minutes that they also have to interpret
Because humans cant think visual. Words have 39 bits of info/s where image is 10
im too poor to see this
this is insane. when tokens are dirt cheap, we can literally do anything. no need to even think about cost-benefit ratios anymore. wild.
@ethanproia insane
probably tokens, but i think it’s a great idea and we’ll definitely see it in the near future
this is sick - not sure if it's efficient
is human thinking visual?
So cool! @_madebygray
woah!!
why would it be ?
I'd want to click the cube and see the email behind that choice.
this visual looks cool! @kylon_ai can you create a version for cloud and birds?
absolute waste of tokens
Consumes more power and slows down the work
How about triangulating?
Very creative! Someone should make a DeepSeek harness plug like this.
Love this idea!!
Love this!
Low iq activities
Look cool but unless you are running long agent requests which generally ppl don’t it’s not useful
Wasted compute visual wouldn't existif humans didn't but uknew that
your linkedin redirection link isnt working on your portfolio!
Tokens go brrrr
Yes please. Reading walls of agent logs is the worst part of building with AI.
devs
Tbh, unnecessary, gotta save that memory for the background processing and tasks.
waste of token..
Is this actually real or just a demo with the images supplied manually? This is amazing, would be great in the future when super small models can do it.
Because text tokens are cheap and fast, while live visual reasoning turns the GPU cluster into a space heater. It's a pipeline cost problem
how do you keep the visual in sync when tool results arrive faster than the animation can transition? do you queue each state or morph straight to the latest one?
