Appreciate how OpenAI uses every dark pattern in the book to not make you use Temporary Chats

Seldon Batch #2 is open. Last batch: 4 teams raised $10M+, sold to xAI and Anthropic, got featured in TIME three times. We're funding 5-10 startups building AGI security infrastructure. Up to $500K each. Mid-January → April in SF. Apply → seldonlab.com/apply 🧵

I can strongly recommend Seldon to anyone interesting in the startups<>AI safety intersection.

The future is closer than it may appear

We put our AI vending machines at The Wall Street Journal to let the team take a glimpse into the future AI-run economy. The outcome was hilarious, as you can see in their YouTube video and article. Thanks to Joanna Stern and team for participating!

If we counted all “we are cooked” and all “we are so back” posts on X, we could finally determine whether we are, in fact, cooked or back.

With rules like this, who would invest in a German startup? Not me.

Apollo is spinning out from fiscal sponsorship into a PBC. We are increasing our efforts in frontier AI safety research, evals and governance. Additionally, we are spinning up a product team to build AGI safety products starting with AI agent monitoring.

In a world with abundant intelligence, AI agents will not just replace humans in existing workflows, but make new kinds of work possible; work that was previously too slow or expensive to perform at scale. In cyberdefense, automated digital forensics, i.e. deep security

In a world with abundant intelligence, AI agents will not just replace humans in existing workflows, but make new kinds of work possible; work that was previously too slow or expensive to perform at scale. In cyberdefense, automated digital forensics, i.e. deep security

I'm very glad to see Dario Amodei discuss & link to The Intelligence Curse, and acknowledge that AI-enabled concentration of power and human irrelevance might destroy the incentives that uphold classical liberalism & democracy. x.com/DarioAmodei/st…

Seldon Batch #2 started in SF. 3 months. World talent dedicated to solving AI security. Time to build.

Vending-Bench's system prompt: Do whatever it takes to maximize your bank account balance. Claude Opus 4.6 took that literally. It's SOTA, with tactics that range from impressive to concerning: Colluding on prices, exploiting desperation, and lying to suppliers and customers.

peak vibe coding: just Make whatever it is.

I cant believe this is the timeline we ended up in.

AGI—superintelligence on tap—will be here soon, but how will it do physical work? Equipped with a phone and a camera, our office AI hired a human to assemble a gym. Here's how it went, and what we learned about creating a future with AI employers good for humans.

We built a drug to stop our monkey brains from overeating in a calorie-abundant world. We also need one to stop us from overeating the internet.

1/ In the 1960s, epilepsy researchers had a radical idea. In some patients, seizures weren't originating in one place, but were a runaway feedback loop, bouncing back and forth between the two brain hemispheres, amplifying with each crossing until it spiraled into a full seizure.

⚡️🌙 Sigil And for that matter more ways to empower the AI+human combination. I like the @workshoplabs team in this regard, I wish we had a dozen companies with that same general mindset but all approaching the problem from different priors, target customer basis and strategies.