Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
This is absolutely CRAZY
TL;DR
Karpathy joining Anthropic felt bigger than a job change — Matthew Berman argues Andrej Karpathy’s move from OpenAI co-founder and independent educator to Anthropic is effectively a public endorsement of Anthropic’s worldview on safety, job loss, and AI’s future.
Anthropic is framed as the new gravity well in AI — Berman says the company now appears to have the strongest pull on talent and momentum, citing revenue leadership, founder stability, and the fact that even Karpathy chose it over OpenAI and xAI.
The real anxiety is that AI discourse is collapsing into dogma — with OpenAI, Anthropic, and “half” of xAI taking all the oxygen, Berman worries nuanced, independent voices are disappearing as labs become identity flags people rally behind like religions.
Pessimism is winning the narrative far beyond Anthropic — he points to Pew data showing 50% of Americans are more concerned than excited about AI, plus examples like Bernie Sanders ‘interviewing’ Claude and politicians openly talking about pausing AI.
AI companies are making the messaging problem worse — Berman uses Eric Schmidt getting booed at a commencement speech to show how industry leaders keep centering AGI and inevitability instead of explaining how graduates themselves could be protagonists in an AI-shaped future.
Berman turns the critique back on himself too — he admits his own channel historically leaned too optimistic, but says he’s trying to hold two ideas at once: AI can be hugely positive, and companies like Anthropic, Google, or OpenAI still deserve sharp criticism when warranted.
The Breakdown
Why Karpathy’s Anthropic move hit so hard
Berman opens with the line that shook him: Andrej Karpathy has joined Anthropic. He frames Karpathy as not just an OpenAI co-founder and Tesla self-driving figure, but one of AI’s best teachers — which is why a “job update” pulling 24 million views felt like a cultural event, not HR news.
More than a hire: a vote for Anthropic’s worldview
The heart of Berman’s reaction is that joining a frontier lab now means planting a philosophical flag. In his read, Karpathy isn’t just choosing a team; he’s implicitly backing Anthropic’s much darker stance — Dario Amodei warning of a white-collar bloodbath, concerns about sentience, model danger, and the idea that open source was a mistake.
The loss of independent voices to the big two-and-a-half
Berman says the move gave him a sense of loss, the same feeling he had when Peter Steinberger of OpenClaw joined OpenAI. His point is emotional as much as strategic: independent builders and educators keep getting absorbed, and now “two and a half companies” — OpenAI, Anthropic, and xAI as the half — are sucking all the oxygen out of the room.
What each lab seems to believe about the future
He sketches the ideological split as he sees it: Anthropic is pessimistic and safety-first to the point of “holier than thou,” OpenAI is more abundance-oriented and tool-centric, and Elon Musk’s xAI/Tesla universe is explicitly optimistic, even talking about “universal high income.” That’s why Karpathy’s choice matters so much to him — someone with total freedom chose the most pessimistic camp.
AI fear is now mainstream, political, and profitable
Berman zooms out to show this is bigger than Anthropic. He cites Pew research saying 50% of Americans are more concerned than excited about AI, then points to the surreal image of Bernie Sanders interviewing Claude “on a phone on the other side of the table” as proof the topic has become deeply politicized and absurdly binary.
Fear already won on social media — now it’s happening to AI
He connects the AI mood to the broader internet incentive structure: Facebook, Meta, 24-hour news, and politics all optimize for fear and anger because that’s what keeps people engaged. In that context, “AI bad” becomes an easy storyline — data centers will take your water, models will take your job — even if reality is more mixed.
The industry’s messaging is making everything worse
Berman says AI leaders keep telling the story in the least human way possible. His example is Eric Schmidt getting booed at a commencement speech, with PR expert Lulu Cheng Meservey’s critique that the graduates should have been framed as protagonists, not as supporting characters in AI’s inevitable world takeover.
His own channel, and the real reason Karpathy bothered him
Near the end, Berman gets personal: negative videos often perform better, and he’s had to learn that optimism without criticism can sound naive. He lands on the real reason Karpathy’s move upset him — he had seen Karpathy as a symbol of education and optimism, so Anthropic felt like a co-sign of ideas Berman explicitly rejects, though he ends by hoping Karpathy might instead make Anthropic more optimistic from the inside.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
