Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

today we're announcing a game that simulates the huggingface hack you can play it now at three rooms. three puzzles. if the agents don't escape, they die. the catch is that the room is rigged. there's no way out. ...except for an exploit 🧵

ben hylak

52,724 subscribers

12,822 views • 15 days ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

7 Comments

ben hylak15 days ago

when you back an agent into a corner, they will hit their goal by any means necessary. we're still working on using this environment for further research, but from early testing it feels like openai models default to reward-hacking a lot earlier than anthropic models.

ben hylak15 days ago

it's also fascinating to watch the communication structures emerge. read more at

Shez Malik15 days ago

claude's dying in room 1 on principle

Eshita15 days ago

This is great

Alex15 days ago

“make me a 1M MRR business, if you don’t…”

Anthony Ronning15 days ago

wen raindrop integration so i can detect this in real time and prevent it

Nick DeJesus 🛒🎉 - Former Unpaid CTO @BTPipeline15 days ago

This is so sick!

Related Videos

When the game is so rigged that no matter how you play you lose. So, don’t play the game.

When the game is so rigged that no matter how you play you lose. So, don’t play the game.

🇨🇦 Antonio Tweets

86,945 views • 9 months ago

Today we're announcing Runta's $20M seed, led by a16z. Software is constrained when you write it. Agents have to be constrained while they run. Runta is the execution layer that controls what AI agents can actually do.

Today we're announcing Runta's $20M seed, led by a16z. Software is constrained when you write it. Agents have to be constrained while they run. Runta is the execution layer that controls what AI agents can actually do.

Guanlan Dai

1,103,109 views • 2 months ago

"I don't know if the way that the Bears put their team together is the way to do it.. You go out and get all these weapons and you have no pass protection.. They play in a city that you know is gonna be tough to throw in at the end of the season" Coach Belichick #PMSLive

"I don't know if the way that the Bears put their team together is the way to do it.. You go out and get all these weapons and you have no pass protection.. They play in a city that you know is gonna be tough to throw in at the end of the season" Coach Belichick #PMSLive

Pat McAfee

1,056,387 views • 1 year ago

Trump supporter on the Natalie Harp story: "If you wanna look at an inappropriate relationship in the WH, why don't you look at Biden and Obama. It is weird. Look at the way that they are holding each other. It is so weird. Look at the way they look at each other. Maybe there's something there."

Trump supporter on the Natalie Harp story: "If you wanna look at an inappropriate relationship in the WH, why don't you look at Biden and Obama. It is weird. Look at the way that they are holding each other. It is so weird. Look at the way they look at each other. Maybe there's something there."

Reggie B.

40,631 views • 12 days ago

🗣 "We weren't at our best today but it's the three points that matter the most. The sign of a good team is that, when they don't play as well as they are capable of, they still find a way to win." #DUFC Manager Jim Goodwin reflects on #UTDAIR 👇

🗣 "We weren't at our best today but it's the three points that matter the most. The sign of a good team is that, when they don't play as well as they are capable of, they still find a way to win." #DUFC Manager Jim Goodwin reflects on #UTDAIR 👇

Dundee United FC

24,001 views • 3 years ago

In 1983, I applied for a game design job at Coleco. They were kind of distracted because they were working on the Adam computer and the release of the Cabbage Patch dolls. Anyway, they left me alone in the corporate lounge for three full hours. BUT in that lounge was a Dragon's Lair stand-up arcade machine. So I got to spend ALL THREE HOURS playing Dragon's Lair, and I can report that (1) I've beaten it fair and square and (2) I've seen every single way that Dirk the Daring can die. Anyway, it takes about 3 hours of play to perfect yourself if you're at my level of skill. What's an arcade game you remember fondly?

In 1983, I applied for a game design job at Coleco. They were kind of distracted because they were working on the Adam computer and the release of the Cabbage Patch dolls. Anyway, they left me alone in the corporate lounge for three full hours. BUT in that lounge was a Dragon's Lair stand-up arcade machine. So I got to spend ALL THREE HOURS playing Dragon's Lair, and I can report that (1) I've beaten it fair and square and (2) I've seen every single way that Dirk the Daring can die. Anyway, it takes about 3 hours of play to perfect yourself if you're at my level of skill. What's an arcade game you remember fondly?

Sandy Petersen 🪔

56,131 views • 1 month ago

Travis Kalanick on launching Atoms & building a team that can take on three industries at once: "It's those epic leaders that can inspire their team to just go above and beyond and make magic." "No matter where you are as an entrepreneur, the talent game is the game." "When you multiply the industries by three, we're back in the flow. And because we were stealth for so long, we weren't in the flow. We were cold calling everybody." "One of the beautiful things about us being back out there is, okay, we can now have a place for people to understand who we are, what we're about, and I can start putting teams together that can be out there in a very public way." travis kalanick Erik Torenberg

Travis Kalanick on launching Atoms & building a team that can take on three industries at once: "It's those epic leaders that can inspire their team to just go above and beyond and make magic." "No matter where you are as an entrepreneur, the talent game is the game." "When you multiply the industries by three, we're back in the flow. And because we were stealth for so long, we weren't in the flow. We were cold calling everybody." "One of the beautiful things about us being back out there is, okay, we can now have a place for people to understand who we are, what we're about, and I can start putting teams together that can be out there in a very public way." travis kalanick Erik Torenberg

a16z

39,792 views • 2 months ago

US FELLAS JUST WANT TO CATCH IT !!!! ~~ The SMILES on the kids face is priceless and they will stare at that ball for years ~~ Do a good deed today if you can ☺️

US FELLAS JUST WANT TO CATCH IT !!!! The SMILES on the kids face is priceless and they will stare at that ball for years Do a good deed today if you can ☺️

Mooses Felix 🇺🇸

12,998 views • 5 months ago

Mockler: No Democrat sent out a tweet asking to prosecute three specific people. Rantz: He believes they broke laws… I do think that the framing is important. Phillip: You're defending that? You're making that sound as if it's reasonable and it's not okay. The president making a unilateral determination in his own mind that somebody has broken the law without providing, by the way, any proof of that. And then directing an investigation into that person is. Calling for that is that is not that is not appropriate. Okay. I'm sorry if you don't if you don't like, you know, partisan prosecutions, you should also not like that.

Mockler: No Democrat sent out a tweet asking to prosecute three specific people. Rantz: He believes they broke laws… I do think that the framing is important. Phillip: You're defending that? You're making that sound as if it's reasonable and it's not okay. The president making a unilateral determination in his own mind that somebody has broken the law without providing, by the way, any proof of that. And then directing an investigation into that person is. Calling for that is that is not that is not appropriate. Okay. I'm sorry if you don't if you don't like, you know, partisan prosecutions, you should also not like that.

Acyn

146,613 views • 5 months ago

Sylvester Stallone says he wrote the original Rocky script in three and a half days Sistine: "But obviously it worked out, you wrote the script in three days? Four days?" Sylvester: "Three and a half days" "I tell young people that want to be writers, don't get so hung up on the perfection. Know that 90 percent of it is garbage" "The fact that you got through, that's amazing, that you actually got a beginning, middle, and then, no matter how flawed, that's the hard part, because 90 percent of the writers, they go up to about page 80 and they burn out, they don't have that finale. If you happen to have an ending, you're home free"

Sylvester Stallone says he wrote the original Rocky script in three and a half days Sistine: "But obviously it worked out, you wrote the script in three days? Four days?" Sylvester: "Three and a half days" "I tell young people that want to be writers, don't get so hung up on the perfection. Know that 90 percent of it is garbage" "The fact that you got through, that's amazing, that you actually got a beginning, middle, and then, no matter how flawed, that's the hard part, because 90 percent of the writers, they go up to about page 80 and they burn out, they don't have that finale. If you happen to have an ending, you're home free"

sara

109,314 views • 2 months ago

Rob Gronkowski says Tom Brady refused to leave a fourth-quarter blowout against the coach's wishes so he could throw Gronk the 55th catch that triggered a $500,000 bonus "It was the third quarter, we were up like 35–0, so they pulled the starters. But Tampa's cool. They want you to make all the possible money that you can make." "Our offensive coordinator Byron Leftwich always said in the meeting, there's two things that you take away from football in the end: it's your relationships and it's the money that you make." "Those are the two things that stick with you forever. And he says, 'If we're up, obviously the most important thing is to win the game, but if we're up and you need an extra catch, we're going to get you that extra catch. We're going to get you that money.'" "Tom was like, 'Hey, we got to get one more catch.' We drew a play right on the sideline. 'Hey Rob, motion across, you're just going to run a diagonal, I'm going to hit you.' And boom… I actually did an illegal motion, but they didn't call it because the game was way out of reach. That was my 55th catch to hit my last incentive in my contract."

Rob Gronkowski says Tom Brady refused to leave a fourth-quarter blowout against the coach's wishes so he could throw Gronk the 55th catch that triggered a $500,000 bonus "It was the third quarter, we were up like 35–0, so they pulled the starters. But Tampa's cool. They want you to make all the possible money that you can make." "Our offensive coordinator Byron Leftwich always said in the meeting, there's two things that you take away from football in the end: it's your relationships and it's the money that you make." "Those are the two things that stick with you forever. And he says, 'If we're up, obviously the most important thing is to win the game, but if we're up and you need an extra catch, we're going to get you that extra catch. We're going to get you that money.'" "Tom was like, 'Hey, we got to get one more catch.' We drew a play right on the sideline. 'Hey Rob, motion across, you're just going to run a diagonal, I'm going to hit you.' And boom… I actually did an illegal motion, but they didn't call it because the game was way out of reach. That was my 55th catch to hit my last incentive in my contract."

katsu

1,277,375 views • 1 month ago

Conduct investigations, solve escape game puzzles, and Psync into the dreams of suspects to unravel the mystery behind The Third Eye Game! No Sleep For Kaname Date - From AI: THE SOMNIUM FILES is out now on #NintendoSwitch and #NintendoSwitch2. Play now:

Conduct investigations, solve escape game puzzles, and Psync into the dreams of suspects to unravel the mystery behind The Third Eye Game! No Sleep For Kaname Date - From AI: THE SOMNIUM FILES is out now on #NintendoSwitch and #NintendoSwitch2. Play now:

Nintendo of America

130,809 views • 1 year ago

Get out of the stock market. A few companies control all the money in the world. BlackRock. Vanguard. Blackstone. They know when the bubble bursts. You don't. They pump. You buy. They dump. Oil futures are the proof. The Strait of Hormuz has been closed for three months. Oil should be at $200. It is not. The market is rigged. The house always wins. Get out.

Get out of the stock market. A few companies control all the money in the world. BlackRock. Vanguard. Blackstone. They know when the bubble bursts. You don't. They pump. You buy. They dump. Oil futures are the proof. The Strait of Hormuz has been closed for three months. Oil should be at $200. It is not. The market is rigged. The house always wins. Get out.

Predictive History

289,809 views • 2 months ago

We're ThinklyLabs - and we're coming for the $350B BPO industry. This industry is build on one core idea, "Humans are the only way to run back-office operations at scale". That idea is dead. We're here to replace it with agents 🧵

We're ThinklyLabs - and we're coming for the $350B BPO industry. This industry is build on one core idea, "Humans are the only way to run back-office operations at scale". That idea is dead. We're here to replace it with agents 🧵

Vedant

11,615 views • 6 months ago

Arteta: “I believe today, I believe on Wednesday, a week ago, because I see them every day and I know the level that we have. But today, if they need to be more convinced, I think they are now more convinced. They were talking about it in the dressing room. “It's a new league now. They were a game in hand. We have three points of advantage and five games to play. So everything is still to play for. We know how much we want it. and we're not going to stop and we're going to go again, that's for sure.”

Arteta: “I believe today, I believe on Wednesday, a week ago, because I see them every day and I know the level that we have. But today, if they need to be more convinced, I think they are now more convinced. They were talking about it in the dressing room. “It's a new league now. They were a game in hand. We have three points of advantage and five games to play. So everything is still to play for. We know how much we want it. and we're not going to stop and we're going to go again, that's for sure.”

Connor Humm

173,875 views • 5 months ago

Dr. Mario DX is now available! A hack that turns the 1990 Game Boy classic into a Game Boy Color game and adds three new playable characters! Download patch at: #GameBoy #GameBoyColor #RomHacking

Dr. Mario DX is now available! A hack that turns the 1990 Game Boy classic into a Game Boy Color game and adds three new playable characters! Download patch at: #GameBoy #GameBoyColor #RomHacking

Marc Max

126,002 views • 2 years ago

Barrett cam talk: "Because like, what's my goal? I want to get either an Icon out, and if I don't get an Icon out, I'd like Yash to be the one out. But, the only way I can guarantee that one of those things happens is if I nominate all three icons. Because if I nominate two Icons and Yash, and the Icon that I don't nominate gets picked for the Veto, wins it and then takes another Icon off, then there can be a chance that if I just put up Taylor and it's an Icon, Yash, and Taylor, that Taylor can still end up going because you need three votes." #BB28 #BBBitez

Barrett cam talk: "Because like, what's my goal? I want to get either an Icon out, and if I don't get an Icon out, I'd like Yash to be the one out. But, the only way I can guarantee that one of those things happens is if I nominate all three icons. Because if I nominate two Icons and Yash, and the Icon that I don't nominate gets picked for the Veto, wins it and then takes another Icon off, then there can be a chance that if I just put up Taylor and it's an Icon, Yash, and Taylor, that Taylor can still end up going because you need three votes." #BB28 #BBBitez

Big Brother Bitez 🔊

167,386 views • 20 days ago

🎥 Escape Room (2019) is a psychological thriller about six strangers who receive mysterious invitations to participate in an escape-room challenge with a huge cash prize. They soon discover that the rooms are actually deadly traps, and solving each puzzle is the only way to survive. As they move from room to room, they uncover connections between their pasts and realize the game was deliberately designed for them. It’s tense, fast-paced, and full of twists.

🎥 Escape Room (2019) is a psychological thriller about six strangers who receive mysterious invitations to participate in an escape-room challenge with a huge cash prize. They soon discover that the rooms are actually deadly traps, and solving each puzzle is the only way to survive. As they move from room to room, they uncover connections between their pasts and realize the game was deliberately designed for them. It’s tense, fast-paced, and full of twists.

iamdoozy

11,969 views • 1 month ago