Replying to @max_paperclips
partly yeah, im so tired that we always say the models are jagged, they are not
training data is jagged and modern neural nets can't generalize well, therefore that's why they are jagged
Replying to @ryunuck
man are you going through a genuine episode of some kind?
i do not want to come off as mean but after reading your posts i feel like you definitely have some kind of mental health problem that needs to be taken care of
please seek help if you need to
Replying to @ryunuck
seriously this has to be some of the dumbest takes i have seen, it's to a point where I genuinely can't tell if you're just ragebaiting.
He quite literally built AlexNet, he invented modern deep learning.
He then jumped to Google and invented Seq2Seq, then was one of the biggest contributors to AlphaGo.
He was also one of the first people when he was at OpenAI to realize and figure out the scaling hypothesis.
Then before he left OpenAI, he was arguably the guy that invented CoT / test time compute for LLMs.
He basically have invented AI as we know it.
If you can name what you disagree / find dumb about him im happy to explain / debate more
Replying to @scaling01
@scaling01 do you expect the open source vs closed gap to get like significantly bigger as the big labs put like their multi million GPU clusters to work or is it still going to be the 6-9 month gap regardless
Replying to @scaling01
@scaling01 do you estimate as of right now is anthropic or openai in lead in terms of internal models
and who are you more bullish on long term
tbh I think I am much bullish on ant long term
Replying to @mboverell
how many X is the max subscription can you please confirm? I am assuming it is 5x weekly then
Replying to @scaling01
@scaling01 bc you have have some special friends and know the models better than me
what's your estimate of ECI of Anthropic and when's your timeline of full RSI (month + year)?
also curious what year do you think we will see wildely superintelligent models? like a model that ex: produces a drug that magically cures alzheimer's that just no human scientist understand at all
Replying to @jparkjmc @hillclimbai
please test opus 5.5 im begging
Replying to @__tinygrad__
@__tinygrad__ thoughts on the Nvidia GPU unlocks?
In the future you will own everything and be happy
Replying to @synthwavedd
is there something that happended within ant or no
Replying to @scaling01
@scaling01 since you have some frontier lab friends, ive always wondered, how true is the secret quantization of openai esp ant models
is it actually happening or just mass hysteria
Replying to @scaling01
@scaling01 my god your take has aged like fine sour milk
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
Replying to @scaling01
I’ve heard a rumour that OAI employees can get like around a few thousand GPUs each just on demand not true how true it is tho
Replying to @juli_li_
when will u guys run opus 5.5 and sol 6.1?
Replying to @NateDBusch @scaling01
agree tbh I don't really believe in the argument that "CoT will work forever"
CoT is going to become neuralease with the next generation of models, just look at the CoT from Fable, no human can read it, they literally need another instance of Fable to read the CoT and translate it into English
Replying to @scaling01
@scaling01 do you think anthropic will do looping models anytime soon or are they too safety pilled to do looping ones
Replying to @jparkjmc
Uhh when will u test opus 5.5 and sol 6.1
Replying to @RhysSullivan
how inspirational! if only you can decompile vscode and open source it for us





