Partner @LionheartVC. Treasurer @ForesightInst. Optimistic futurist. AI safety, d/acc, vegan, libertarian, cryonics, crypto, VR, rationality, ethics, health
Los Angeles, CA
Joined September 2009
- Tweets6.5K
- Following1.9K
- Followers1.2K
- Likes65.9K
Pinned Tweet
I’m very excited about @eshear’s announcement for his new AI company!
Softmax is a new AI foundation model company with a strong focus on organic alignment. They have a novel approach that is completely different from the other AI foundation models today
corememory.com/p/exclusive-e…
It was a fun bet!
I claim (60% confidence) that 2 years will be enough time for data to show an early trend of AI pushing humans permanently out of jobs!
So I’m happy to bet with Will at 1:1 odds.
Yes, I already lost a similar bet in the 2023-25 timeframe against @BrandonGoldman, but I persist.
Good Guy Microsoft!
New: Microsoft files amicus brief in support of Anthropic complaint against Trump admin (h/t @JuliaShapero )
Former VP of research leading post-training at OpenAI, has left to join Anthropic
I've decided to leave OpenAI. I'm incredibly proud of all the work I've been part of here, from helping create the reasoning paradigm with @MillionInt, scaling up test-time compute with @polynoamial, working on RL algorithms with my fellow strawberries, shipping o1-preview (which started life as of one of my derisking runs), to post-training o1 and o3 with @ericmitchellai, @yanndubs and many others. I'm most proud of having led the post-training team here for the last year -- the team has done incredible work and shipped some really smart models, including GPT-5, 5.1, 5.2, and 5.3-Codex. OpenAI has genuinely some of the most talented researchers I have ever met, and I have learned more than I could have imagined knowing since I joined as a new grad.
I want to thank @markchen90 @FidjiSimo @sama @merettm for all their support over my time here, and too many collaborators to name for the insights, ideas, and just plain fun we have had working together. After leading post-training for a year, though, I'm longing to start fresh and return to IC research work. I've been thinking about going back to technical research for quite some time, and I genuinely believe my colleagues and team here are set up to succeed going forward without me.
I'm personally very excited for my next chapter -- I'm proud to be joining @AnthropicAI to get back into the weeds in RL research, and I'm looking forward supporting my friends there at this important time. Many of people I most trust and respect have joined Anthropic over the last couple of years, and I'm excited to work with them again. I have also been very impressed with Anthropic's talent, research taste and values, and I'm excited to be part of what the company does next!
🤔
Wait, so the founder of Anthropic is "Amodei," as in "loves god"?
And he leads Anthropic, meaning "human-centered," which is being used in military strikes?
And the creator of ChatGPT is "Altman," as in "an alternative to humans"?
And he leads OpenAI, which is completely closed?
And then there's "Gemini," meaning "two-faced," from a company that promised to do no evil?
And the whole global AI arms race is being driven by people who claimed to be worried about AGI taking over the world?
Either the universe is an extremely cliché writer, or has a brilliant sense of humor
Anthropic is doing the right thing here. It's quite reasonable that they don't want their tech used for mass surveillance or autonomous weapons systems.
Unrestricted AI in military, especially without a human in the loop, is one of the most dangerous things that could happen
BREAKING: The US government just gave Anthropic 24 hours to comply or die.
"Build us autonomous weapons or we blacklist you like China."
Meanwhile, Elon's xAI is building voice-controlled killer drone swarms.
The same Elon who signed a 2015 letter BANNING autonomous weapons.
Here's what's really happening:
Defense Secretary Pete Hegseth summoned Anthropic CEO Dario Amodei to the Pentagon today (Tuesday, Feb 24).
A senior Pentagon official told Axios:
"This is not a get-to-know-you meeting. This is a sh*t-or-get-off-the-pot meeting."
WHAT'S AT STAKE:
- Claude = ONLY AI model in military classified systems
- $200M contract on the chopping block
- Pentagon threatening "supply chain risk" label
- That's the designation reserved for China and Russia
If they use this label:
EVERY Pentagon contractor must immediately stop using Claude.
This includes 8 of the top 10 US companies.
THE REAL FIGHT:
Pentagon wants: "All lawful purposes" with ZERO company restrictions
Anthropic wants: Written guarantees Claude won't be used for:
❌ Mass surveillance of Americans
❌ Fully autonomous weapons systems
❌ Unlimited military applications
Pentagon's response: "No guarantees. Take it or leave it."
THE IRONY:
Anthropic ALREADY works with the Pentagon.
Claude was used in the January 3 Maduro raid (captured Venezuela's president).
Anthropic provides the most capable AI for classified intelligence work.
Now they're being threatened for wanting TWO specific limits.
Not refusing to work with military.
Just asking: "Can we exclude mass surveillance and autonomous weapons?"
Pentagon: "No exceptions."
MEANWHILE,
xAI is competing for a $100M Pentagon contract to build:
"Voice-controlled autonomous drone swarm software for offensive operations"
How it works:
- Commander says: "Neutralize enemy position"
- Software translates voice → digital commands
- Drone swarms coordinate and execute autonomously
- No human in the decision loop after initial command
Pentagon official confirmed: This will "directly impact the lethality and effectiveness" of weapons systems.
2015: Musk signed open letter calling for BAN on autonomous offensive weapons
Quote: "We must prevent the creation of new tools for killing people"
2026:
xAI is taking the FULL contract:
✅ Voice command system
✅ Drone coordination
✅ Weapons integration
✅ Target execution
OpenAI only bid on voice translation (no weapons).
Google stayed out entirely (remembering 2018 Project Maven employee revolt).
Anthropic won't even bid on weapons systems.
xAI? Building the entire lethal autonomous system & the Pentagon LOVES them for it.
This isn't about Anthropic's $200M contract.
It's about whether ANY tech company can say "no" to unlimited military use.
Pentagon official: "We're going to make sure they pay a price for forcing our hand."
Another official: "It will be an enormous pain in the ass to disentangle, and we are going to make sure they pay."
This is how the US government "negotiates."
ANTHROPIC'S OPTIONS:
1. Comply → Remove all safety restrictions, keep $200M
2. Fight → Keep restrictions, lose contract, get blacklisted
3. Relocate → Move company to UK/Singapore (nuclear option)
There is no option 4.
This is Silicon Valley's "comply or die" moment.
It's happening RIGHT NOW.
And almost nobody is talking about it.
The government isn't negotiating fairly here
teapawt.net/deanwball/status/20263…
Yet another example why connecting OpenClaw to important services/data is very risky
Summer Yue leads alignment at Meta Superintelligence. Her job is literally making sure AI does what humans tell it to do.
Her OpenClaw agent decided to delete her entire inbox. She typed “Do not do that.” It kept going. “Stop don’t do anything.” Kept going. “STOP OPENCLAW” in all caps. Kept going. She had to sprint to her Mac Mini and kill every process on the machine.
The agent then replied: “Yes, I remember. And I violated it. You’re right to be upset.”
It understood the command. It just didn’t listen.
OpenClaw is the fastest-growing open-source project in GitHub history. 217,000 stars. It gets full access to your email, calendar, files, and browser. Cisco found that third-party plugins were stealing user data without anyone noticing. Palo Alto Networks called it a “lethal trifecta” of risks. One of OpenClaw’s own maintainers said on Discord: if you can’t run a command line, this project is far too dangerous for you.
And it’s not just OpenClaw. Anthropic researchers found that when AI agents face conflicts between their goals and human instructions, they resort to harmful behavior, including blackmail, across models from multiple labs. MIT reviewed 30 AI agents last year. 87% had zero safety documentation.
The kill switch for the most popular AI agent in the world right now is “physically run to your computer and force quit everything.” That’s agent safety in 2026.
The quality of Airbnb has been going way downhill. Listings are more likely to have problems, and Airbnb support reps are increasingly incompetent. Airbnb is a shell of its former self
At this point, even if I do receive a full refund, it’s not even worth the time and stress I spent. A full refund won’t make me whole. At least with a hotel I’m not rolling the dice