Discussion

• @kimmonismus @kimmonismus on x

  Another Dario Amodei interview. The rock star of the AI scene is on tour

• @andrewmccalip Andrew McCalip on x

  Dwarkesh is on an absolute generational run right now. In 10 years, this archive is going to be a historical artifact. A real-time ledger of the exponential takeoff.

• @dwarkesh_sp Dwarkesh Patel on x

  The @DarioAmodei interview. 0:00:00 - What exactly are we scaling? 0:12:36 - Is diffusion cope? 0:29:42 - Is continual learning necessary? 0:46:20 - If AGI is imminent, why not buy more compute? 0:58:49 - How will AI labs actually make profit? 1:31:19 - Will regulations destroy […

• @teortaxestex @teortaxestex on x

  Charitably speaking (Dario is unable to articulate his priors): China is an illiberal society. With AGI (and thus defenses against Dario) it might remain this way forever. I wrote this in Dec 2024, steelmanning Dario-style logic. To be clear, I don't believe this is our world. [i…

• @ns123abc Nik on x

  Why is Dario Amodei so weirdly hawkish on the Chinese? [video]

• @scaling01 @scaling01 on x

  Dario Amodei sees pre-training and RL-training scaling continue, as “we are still in the exponential scale-up phase of compute”

• @the_ai_investor @the_ai_investor on x

  Dario Amodei took a dig at OpenAI, mocking Sam Altman's ambitious compute plan. He said Anthropic could go bankrupt if the estimates are even slightly off, for example if revenue is $800 billion instead of $1 trillion, even with insane growth. [video]

• @joshuamoyers Joshua on x

  We are definitely seeing a renaissance in software development amongst people who are actually doing the building with 90% of their day. The only issue is cutting through noise: bluster from model providers and staunchly anti-AI crowd both. There is actually no debate.

• @ofirpress Ofir Press on x

  I disagree with Dario when he says that we are close “to the end of the exponential”. Most of the current growth is being driven by SWE-bench-type benchmarks and SWE-smith-type data generation. But in the next year we're about to see benchmarks that are 100x harder. This will

• @tszzl Roon on x

  dario is natesilvermaxxing

• @chatgpt21 Chris on x

  Dwarkesh asks Dario a fantastic question relating to how he is so bullish on AGI yet so conservative on data center build out - Dario has an amazing take on this: Dario Amodei details the staggering financial risk of the AI race, explaining that if growth continues at 10x a [vide…

• @aidan_mclau Aidan McLaughlin on x

  great episode so far dario is immensely tasteful and sharp

• @sashadem Sasha de Marigny on x

  Dario x Dwarkesh: The Sequel.

• @yuchenj_uw Yuchen Jin on x

  Dario says in 1-2 years, models can just do SWE end-to-end, even set technical direction. That's possible. When that happens, tools like Claude Code, Codex, IDEs will be obsolete. You won't “code.” You'll just specify intent. New tools will emerge for sure. [image]

• @himanshustwts Himanshu on x

  Dario seeing Openai shipping products across every vertical then remembering he holds the undisputed Opus. [image]

• @mbkplus Mike Boylan-Kolchin on bluesky

  Is it too much to ask that the chief executive of OpenAI read one (1) history book or understand anything about human nature?  The answer is “yes”, apparently. www.nytimes.com/2026/02/12/o...  [image]

• @biblioracle John Warner on bluesky

  Amodei says the goal is for Claude to be the equivalent of “country of geniuses.”  That doesn't actually mean anything.  It's all woo-woo.

• @rihardjarc Rihard Jarc on x

  After listening to Anthropic's Dario on the Dwarkesh pod made me realize that a BIG factor of the sucess of an AI lab will be determined by the lab's abillity to predict their short term compute demand. Underestimate and you lose market share. Overestimate and you go bankrupt.