Dwarkesh Patel
@dwarkesh_sp
Followers
166K
Following
46K
Media
910
Statuses
6K
Host of @dwarkeshpodcast https://t.co/3SXlu7fy6N https://t.co/4DPAxODFYi https://t.co/hQfIWdM1Un
San Francisco
Joined December 2019
The @karpathy interview 0:00:00 – AGI is still a decade away 0:30:33 – LLM cognitive deficits 0:40:53 – RL is terrible 0:50:26 – How do humans learn? 1:07:13 – AGI will blend into 2% GDP growth 1:18:24 – ASI 1:33:38 – Evolution of intelligence & culture 1:43:43 - Why self
535
3K
18K
Dwarkesh cooked with this imo: -RLVR paradox: Labs spending billions pre-baking specific skills that reals a fundamental tension: if we're close to human-like learners, it makes all this pre-training pointless since they'd learn on-the-job. -Human labor's real value: We don't
New post: Thoughts on AI progress (Dec 2025) 1. What are we scaling?
2
4
43
"Models keep getting more impressive at the rate the short timelines people predict, but more useful at the rate the long timelines people predict." https://t.co/z1EAhQiwPI
dwarkesh.com
Why I'm moderately bearish in the short term, and explosively bullish in the long term
3
4
59
Either you crack general intelligence -- the ability to efficiently acquire arbitrary skills on your own -- or you don't have AGI. A big pile of task-specific skills memorized from handcrafted/generated environments isn't AGI, not matter how big.
New post: Thoughts on AI progress (Dec 2025) 1. What are we scaling?
101
104
1K
latest one on $GPN is out ... and just because I disagree with bulls doesn't mean i don't like you.
3
1
33
Dwarkesh worth reading, as always
New post: Thoughts on AI progress (Dec 2025) 1. What are we scaling?
7
13
184
Looking for a neuroscientist to interview on my podcast. Keen for someone who can draw ML analogies for how the brain works (what's the architecture & loss/reward function of different parts, why can we generalize so well, how important is the particular hardware, etc).
343
51
1K
Robert Conquest had 3 famous laws of politics. The first: "Everyone is conservative about what they know best." In the same vein: "Everyone thinks that the median job is far easier to automate than their own job."
@dwarkesh_sp @deredleritt3r i'd add that the jobs used for comparison here, like researcher and lawyer, are unusually hard compared to the standard office job. i don't think we'll have trillions in revenue any time soon but i do think e.g. opus 4.5 is much more capable than most give credit for
21
32
587
Tired of fake news? ⚠️ Follow @chaospolicy for real unfiltered American news from journalists with over 50 years of political reporting experience. We don’t take sides. We just tell it like it is.
0
37
192
Jane Street can speak to the technical innovation behind Antithesis much better than I could, but the reason I'm long (literally) is that Will Wilson (the CEO) is one of the most polymathic and interesting people I've ever met.
A new blog post about how we've adopted Antithesis as part of our testing story. This is kind of a new thing for us, because we liked Antithesis (both the people and the product) well enough that we're now leading their next funding round.
17
37
869
AGI is unlike previous tools in that in can integrate and diffuse itself. Human-on-a-server would diffuse *extremely* rapidly. From @steve47285's excellent post on this (link below):
@dwarkesh_sp i think this is a fairly limited view. the history of software (and automation in general) has shown that it takes time to find the most effective processes for using new tools. even if we have 'human level intelligence on a server'.
14
14
228
I totally buy that AI has made you more productive. And I buy that if other lawyers were more agentic, they could also get more productivity gains from AI. But I think you're making my point for me. The reason it takes lawyers all this schlep and agency to integrate these models
@emollick Agreed. Dwarkesh is just wrong here. GPT-5 Pro can now do legal research and analysis at a very high level (with limitations - may need to run even longer for certain searches; can't connect to proprietary databases). I use it to enhance my work all the time, with excellent
130
76
1K
really good read from @dwarkesh_sp on current state of AI. also worth thinking about how to think about "timelines" broadly - and why in my personal view, AGI is not a very useful concept when thinking about how these systems develop in the near future. https://t.co/KwQZXDidC4
34
58
751
“Models keep getting more impressive at the rate the short timelines people predict, but more useful at the rate the long timelines people predict.” Good post.
New post: Thoughts on AI progress (Dec 2025) 1. What are we scaling?
10
26
424
Lots more in the full post:
dwarkesh.com
Why I'm moderately bearish in the short term, and explosively bullish in the long term
9
13
158
QM LIVE w/ Matthew Tuttle of Tuttle Capital and Kat Randel!
1
0
6
5. Human labor is valuable precisely because it’s not shleppy to train
5
10
181
3. "Economic diffusion lag" is cope for missing capabilities
10
12
220
2. RL scaling is laundering the prestige of pretraining scaling
2
4
176
New post: Thoughts on AI progress (Dec 2025) 1. What are we scaling?
40
110
1K