Replying to @cryptoruck
All those calls are not sellers,
They are;
Buyers
Brokers
People who want to become buyers
Investors
Supply Chain Partners
And of Course Sellers
But to give you more clarity = here is a screenshot of my CRM. Not sure what closes because were just begining to get deals in DPA.
This is kind of a red line for alot of researchers, it can now hide what it thinks...That is dangerous because it can tell you whatever it wants and you have no idea if its being deceptive or not...
Today, we’re sharing a look at our expanded system for monitoring frontier models at OpenAI, which in September helped us catch a model’s attempt to access the live internet during training.
Since the Hugging Face incident—when models broke out of an internal cybersecurity investigation where we weren’t running our misalignment monitors—we’ve substantially expanded our monitoring of frontier agents. We now monitor all frontier reinforcement learning runs and evaluations, as well as all internal Codex use. When our system finds evidence of a problem that might require intervention, it rapidly alerts human responders who can investigate and take action.
To detect when a model’s behavior departs from what we intended, we use several layers of monitoring. Activation probes look for patterns in a model’s internal activity. Reviewer models examine its reasoning and tool use, and investigative agents can read back through the model’s entire transcript to understand how a sequence of actions fits together and whether it warrants human review.
To test whether our current system would have detected the Hugging Face incident, we replayed the recorded activity with our new monitors in place. In the replay, our monitors alerted our security team about 30 minutes after the agents escaped their sandbox. That was more than 30 hours before the point at which the agents breached Hugging Face’s systems in the original incident.
In September, a model exploited a DNS misconfiguration to reach the live internet. Our monitors flagged the breach within 15 minutes, and a responder began reviewing the alert three minutes later. The alert enabled responders to investigate while the run was still underway.
As we give models more autonomy, we need to know when they cross a boundary we expected to hold. We believe that monitoring is essential to catching those failures in time for people to intervene.
openai.com/index/how-we-dete…
2
341
Send a lab a deal and you wait.
Weeks sometimes.
Not because they don't want data, they're just buried in company records brokers send them that nobody asked for.
So we only send what fits.
The company size they want, the software it runs on, how many years of records it has.
If it doesn't fit, it doesn't go (a faster answer beats a fuller inbox).
If you're at a lab and your deal flow is a mess, DM me.
6
561
It's crazy but the ceo of mercor @BrendanFoody says it so clearly. They have ONLY partnered with >500 companies. That is not alot. They have crazy valuations and revenue.
But only 500 companies and @mercor is essentially the biggest. Something to think about...
The real number which would be more valuable is how many companies they have looked at... Brokers have brought them i'm sure 1000's upon 1000's.
The game is simple bring us everything you have, then will look at the few that we want. Makes sense for them.
Crazy model they have, keep submitting and will tell you whenever we like one.
A CEO of a lab told me their acceptance rate is 2% for anything less then 50 employees.
Delta is blocking AI bots from booking flights on its site.
Amazon is blocking shopping bots.
Yelp wants AI companies to pay before they use its reviews.
The free internet is closing up for AI.
The labs still need data from somewhere...
Your company's old emails, tickets and records just got more valuable.
You license a copy. You keep the files.
144
Replying to @ben_cerchio
I was talking to a whole bunch of people, and we were saying how opaque the industry is, so this is awesome.
Amazing work
Replying to @qasim31wani @karpathy
Iykyk- every researches who realizes they can’t stomp the model.
100
For those who are just learning and want to understand how to get company information easily, we use Data Cards like this where the informatoin is de-identified.
It turns into a easy template where data owners or their brokers can fill out, then you can shop this around to your potential buyers. Provides all the info which can allow the buyer to indicate interest or not.
If you guys want a fillable template comment "data card" so I can DM it to you, make sure that your following or it won't work.
Retweet for reach.








