Viaje de 15hs. No tengo a nadie sentado en mi fila ni tampoco en la de atrás (o sea, no hay patadas). No hay ningún bebé a la vista. Todo este karma positivo significa que el avión se va a caer en el medio del atlántico a mi no me engañan.
Me fascina el fenómeno de convergencia fenotípica de Maru botana y Macri con el pasar del tiempo. No tengo dudas que a los 90 años van a ser prácticamente indistinguibles el uno del otro.
Laughing so hard realizing that to properly use an LLM requires you to be a mystic.
I think we're in for a new age of mysticism the STEM never could have considered.
The Pentagon is about to give an American AI company the Huawei treatment.
Not because it’s Chinese. Not because it’s a spy risk.
Because it refuses to let the military use its AI for mass surveillance of Americans and fully autonomous weapons.
This morning, Defense Secretary Pete Hegseth summoned Anthropic CEO Dario Amodei to the Pentagon. A senior Defense official told Axios: “This is not a friendly meeting. This is a sh*t-or-get-off-the-pot meeting.”
Here’s what’s actually happening:
Claude is the only AI model running inside the Pentagon’s classified systems. The most capable model for sensitive defense and intelligence work. It was used in the Maduro raid in January through Palantir, the first confirmed use of a commercial AI in a classified military operation.
Now the Pentagon wants all restrictions removed. “All lawful purposes.” Including capabilities that would let the military continuously monitor the social media posts, voter registration, concealed carry permits, and demonstration records of every American citizen using AI at scale.
Anthropic said no to two things: mass surveillance of Americans and fully autonomous weaponry.
The Pentagon’s response: threatening to designate Anthropic a “supply chain risk.”
That designation is reserved for foreign adversaries. The last company to receive it was Huawei. It would force every defense contractor in America to certify they don’t use Claude in their workflows. Given that 8 of the Fortune 10 use Claude, this would cascade through the entire defense industrial base.
A senior Pentagon official told Axios: “It will be an enormous pain in the ass to disentangle, and we are going to make sure they pay a price for forcing our hand like this.”
Another official: “The problem with Dario is, with him, it’s ideological. We know who we’re dealing with.”
Meanwhile: OpenAI, Google, and xAI have already agreed to remove their safeguards for military use. OpenAI deployed ChatGPT to all 3 million DoD personnel through GenAI. mil. xAI holds a separate $200M contract backed by Musk’s political proximity to the administration.
Anthropic is the only one that said no.
Think about what’s being asked. The company whose own safety chief resigned two weeks ago warning “the world is in peril.” The company that just published a report showing its most advanced model “knowingly assisted with chemical weapons research” in testing. That company is being punished for refusing to hand the U.S. military unrestricted access to that same technology.
The Pentagon admits competing models “are just behind” for classified work. They need Claude. But they’re willing to blow up the relationship rather than accept two restrictions that protect American citizens from their own government.
This is the most important story in AI right now and almost nobody is framing it correctly.
It’s not about one $200M contract. It’s about whether the U.S. military can compel a private company to remove safety restrictions on technology its own developers have demonstrated is dangerous, under threat of receiving the same designation as a Chinese national security threat.
Dario Amodei walks into that meeting this morning with $380 billion in enterprise value, $14 billion in revenue, and a principle that may cost him both.
Full institutional analysis on my Substack.
https://t.co/AEv8EMPdsZ
Como alguien que habita el uso cotidiano de AI para la ingeniería de proteínas para la salud y erradicación de enfermedades, me parece escandaloso lo *en bolas* que estamos algunos de mis colegas y yo a nivel filosófico para debatir sobre el impacto global de lo que se viene.
This is being read as a philosophical farewell. It’s a resignation letter from the head of Anthropic’s Safeguards Research Team, and the most important sentence is buried in paragraph three.
“I’ve repeatedly seen how hard it is to truly let our values govern our actions. I’ve seen this within myself, within the organization, where we constantly face pressures to set aside what matters most.”
That’s the person responsible for keeping Claude safe telling you the pressures to ship are winning.
Mrinank Sharma built the Constitutional Classifiers system, developed defenses against AI-assisted bioterrorism, and authored one of the first AI safety cases ever written. Two years of work at the exact intersection of “make the model safe” and “ship the model fast.” And he just walked away.
Now zoom out. Dylan Scandinaro, another Anthropic AI safety researcher, left last week to become OpenAI’s Head of Preparedness. Harsh Mehta and Behnam Neyshabur, both senior technical staff, also departed in the past two weeks. Four notable exits in a single month from the company that sells itself as the responsible AI lab.
Meanwhile, Anthropic is in talks to raise at a $350B valuation and just launched Opus 4.6 last Thursday. The commercial engine is accelerating. The safety talent is dispersing.
This is the core tension of every AI company right now: the people building the guardrails and the people building the revenue targets occupy the same org chart, but they optimize for different variables. When the pressure to scale wins enough internal battles, the safety people don’t fight forever. They leave and write beautifully worded letters about integrity.
Sharma’s next move tells you everything. He’s pursuing a poetry degree. When your head of safeguards research decides the most authentic use of his time is writing poems instead of writing safety cases, that’s a signal about what he believes the safety cases were actually accomplishing.