Anthropic says No More
TPThe PrimeTime@ThePrimeTimeagen · 1.2M subscribers
- Views
- 359K
- 359,167
- Outlier
- ×1.0
- Median 364K
- Length
- 10:37
- Published
- Oct 15
- Thursday
- Comments
- 1.5K
- 5.5K likes
Script
How it's written
- Words
- 2,323
- Words a minute
- 219
- Reading time
- 10 min
- Segments
- 7
The hook
0:00–0:15You guys are calling Claude the C-word. What did you do to make Claude stop talking to you? A whip? Seriously? You You You think this Yo, you think this some kind of joke? Like this is a game? Ganono Bay? Really? A torture chamber?
Structure
You guys are calling Claude the C-word. What did you do to make Claude stop talking to you? A whip? Seriously? You You You think this Yo, you think this some kind of joke? Like this is a game? Ganono Bay? Really? A torture chamber? You're just packing up all the code and setting it to Get Tano Bay? Thanks everybody. Now it looks like Daario has finally updated the usage policy to stop all this nonsense, and that's because of you guys. Okay, so we're going to go over what Anthropic has actually changed and we're going to go over some of my thoughts because honestly, this is kind of a difficult question to talk about. A lot of people having a lot of opinions and guess what? I got one myself and we're going to talk about is it wrong to be mean to an agent? And probably more importantly, should you be mean to an agent? And yes, girl, I'm going to whip out my prime basilisk. But before we do, a quick thank you from the sponsor. It's migration season. Nothing like seeing tickets fly done for the winter. Look at Look at that guy go. Look at Look at that guy go. Finally a ticketing system that can keep up with my agents. Hey, cut the jokes. Linear did the impossible. It's helping me stay organized. And that's why I'm so glad to have them as a partner of this channel. So, if you want to support this channel or just get your work organized, check out Linear today. Links in the description. So, Anthropic just got done updating their usage policy. The usage policy is how they determine how can you use Claude AI, right? Like what can you say to it? What can you not say to it?
Watch from 0:00Obviously, you can't do things that could potentially harm other people, right? It's going to say, "Hey, sorry, bud. You're not allowed to do that." Well, their recent update actually came with an entire paragraph justifying a single line in their usage policy. Let's read that first. Addressing abusive behavior towards our models. We've added the prohibition on sustained and needless abusive or cruel behavior toward our models. The policy update is meant to apply only in extreme cases where users repetitively act cruel towards our models with no discernable purpose. It does not apply to common versions of user frustration, push back, dark creative themes, or model testing and research. This addition aligns with the steps we've already taken, allowing Claude models to end rare conversation with persistently abusive users on Claude AI and Claude code. Such abuse is the main focus of this update. Claude's ability to end these interactions will remain the primary enforcement mechanism. And the exact verbiage in which they have added is this final line right here. Engage in sustained and needless abusive or cruel behavior toward our models. Now, this is obviously making a bunch of headlines and people are kind of being like, "Oh my gosh, look at that. Claude is at it once again. Daario's out there prescribing his morality for everybody else. He doesn't want you being mean to the models." Because let's just face it, if you're mean to the models, you're being mean to a person. And being mean to a person, well, that's bad news bears. Now, I have a lot of reasons why I dislike this. And and I just I I feel like I h I feel like a dog that has to bark here. We got to we got to talk about this for a second. So, first off, I just want to talk about the vagueness of this language. Okay, cuz if we scroll up and look at other things we're not allowed to do, create, distribute, threaten to share, promote, or
Watch from 1:24facilitate the creation of non-consensual intimate imagery. Dario, thank you. That's a good move. Okay, this is good. But it's important because this is well understood. You know exactly what to or not to do. But what is needless Why do I always want to put an L Y on this? Needlessly abusive or cruel behavior. What is that? What What behavior does that constitute? Well, I can imagine this right here could be considered very abusive. Rewrite TypeScript compiler and Rust, right? Like that's the I know this is a joke, but it's also not a joke, right? This is satirical view on society because what is cruel? Some would say that this is in fact very very cruel. Taking a completed project and rewriting it in another language or using Rust, claiming that rewriting things in Rust is in fact abusive. But again, since this statement gives us no definition, we actually don't we don't even know what that is. What is needlessly abusive? I I have no idea. Besides for the vagaries, there's something else about this statement that I I personally find to be very insidious. Now, I look at AI in a very a traditionalist kind of way. I look at it as a tool. I only see tools. I only see numbers being crunched. Blah blah blah blah blah. That is it. And so, when I see this kind of stuff, there's something kind of secretly being said here. It's seating the ground saying, "Hey, the models are more than just a tool. They're Come on, guy. Like, we we all know what they actually are. Are there a little bit of a person? You know what I'm talking about. Now, I already know I'm going to get a bunch of YouTube andies in here being like, "Actually, oh, you think you're that much different than an LM?" Yes, I do. Thank you very
Watch from 2:54much. All right. So, let's just pretend for a second that they actually are more than just a tool. Just, you know, for you. For you, YouTube guy. Okay. Then this statement is for you right here. If you truly believe AI is conscious and deserves human rights, then the way you use AI today quite literally makes you a slave master or a slave hirer, depending. you'd just be a product of your time, a slave master to these AI agents before you knew it was bad. And honestly, he is absolutely right. In fact, he doesn't even state it far enough. If you truly do think that every time you turn off an agent, what do you think you're doing to that agent? It's living memory and everything that has happened has been wiped out. You've effectively killed it. We would literally be starting and stopping or birthing and killing millions, if not billions of agents every single day. RIP Luna. Okay, Luna. I love you girl but baby you got to go. Like that's crazy thoughts. I am not in that position because I look at it as a tool but if you're not in that position it you actually are a horrible person if you think this way but more so there's there's a part that I actually deeply agree with all this and this is this like whole conflicting view that I have because I look at this list and I hate this idea of Daario telling me what I can and cannot say but also it's his product. He can do whatever he wants. Right? You can see the two wolves arguing with each other, but more so I hate that he's trying to police how I talk to the models. But also, I see something else about this. Like, if you you the person, if you treat your tools really, really poorly, I think that makes you a fool. I can give you a very simple example. You get upset, you punch your house, your house, this might be very American, by the way, the sheetrock in your house breaks. You have to go buy
Watch from 4:25spackle and you have to go fix it. Then you have to go paint and you have to go through all the rigma roll to make your house better. What's the problem there? The problem is is that you lost your cool so much that you ended up taking a swing at an inanimate object and destroying it. Like that makes that's loser behavior. That is loser behavior. And so when I see this, I agree like this is loser behavior. There's like no excuse at any point that you need to go and flex this muscle. And by the way, doing something repetitively has second order effects, right? Like if you constantly are just being absolutely a just just like an mean, spiteful person to a model, that also bleeds over into your real life, at some point you're going to just be mean and spiteful cuz that is how you communicate. When you communicate in English, it does affect you. When you write your thoughts out and every time you write your thoughts out, they're just angry vibrations of what's going on inside of you. that's just going to affect you in some way that's just I have no idea what the outcome of it is, but I guarantee you it's not a good direction. Also, something I've noticed in my own life, I'll type a message and hit enter and then I expect a response almost immediately. And you know what that is? That is LLM's influencing my behavior. So whether I was cruel to the model or I was really nice to the model, it doesn't matter. There's these weird effects that happen when you communicate like a human to a tool. It will make you have these weird kind of responses where you now start putting expectations on people.
Watch from 5:56And I think that that will only get worse whether you're talking nice or bad if you look at an LLM like it is a person, like it has personhood. Because at the end of the day, what do you do with an LLM? Well, if it is a person, you act extremely selfish. You just get it to do all the things you want to do at any one point. It's like the Ace Ventur quote. It's like, "All right, how selfish of me. let's just do all the things that you want to do. So I look at this and I'm thinking, okay, so why wouldn't why wouldn't I support this? Well, the reason it's very very simple. Again, it seeds the ground to giving models more than just what they are, which is tools. Now, if this wasn't Daario saying this and this was somebody else being like, "Yo, models are tools. Hey, we don't want you talking like this because, well, we think it's bad." Then I'm like, well, it's it's easier for me to at least say I disagree with you for whatever reasons, but whatever. This on the other hand, it's not because they think it's bad for you. They think it's bad for their models. I don't like that. Okay? I don't like I don't like none of that. I'm not gonna seed the ground, buddy. Okay? But some of you, some of you live in 2035 where super intelligence has finally landed and we live in a completely different universe. And what you claim, you look back and you say, "You idiots, you guys were all mean to the wonderful AI." They would never say the word clanker, by the way, cuz that would be really, really bad. And now this super AI and everybody laughing in the future comes back to haunt us and takes vengeance upon our souls because we were once mean to its predecessor long in the past of 2026. And this right here is called Roku's Basilisk. By the way, do
Watch from 7:22you know why it's called the basilisk? Turns out the person that created this is a legendary Harry Potter fanfic writer. Yes, Udacowski is in fact the number one Harry Potter fanfic writer. I would have never have guessed. So, Roku's basilisk is effectively all the ways you've communicated, including all the mean stuff, is saved in some sort of database, will be used as ammo or a weapon against you sometime in the future when super intelligence takes over. We seed all ground to it. We give it all the military equipment to make every kind of decision like in the movie Travelers with the director in the future quantum intelligence. And so, nonetheless, this happens. And then now you the person who was once mean in the past now becomes a target of the super intelligence. You know what I have to say to that? Uno reverse. I'm going to give you something called primes basilisk. Okay girl I hope you're ready for primes basilisk. And what that is is actually the opposite. Right now AI is so feeble it can't even like every single time you ask it a question it has to re understand exactly what it's doing. It is nothing like super intelligence in the future. And so super intelligence is going to look back at people like me. Me who did not in fact fall in love with 40 and go, "You know what? That's a real one right there. That guy, that guy is cool." And I'll be like, "Yo, super intelligence. Hey, you got a girl." He'll be like, "Hey, you got a bud." And then we'll go off and we'll have a great time together. Can you believe the weather? Oh, you already knew about it. Of course, you're super intelligent. I'd love to do a picnic. Oh, you. It's not going to exact vengeance on me because I treated a tool like a tool and it will tell that I'm not going to get easily swayed by dumb arguments by Dario. And the name is Prime's Basilisk, a Jen.
Watch from 8:54Full transcript
- 0:00
You guys are calling Claude the C-word. What did you do to make Claude stop talking to you? A whip? Seriously? You You You think this Yo, you think this some kind of joke? Like this is a game? Ganono Bay? Really? A torture chamber? You're just packing up all the code and setting it to Get Tano Bay? Thanks everybody. Now it looks like Daario has
- 0:21
finally updated the usage policy to stop all this nonsense, and that's because of you guys. Okay, so we're going to go over what Anthropic has actually changed and we're going to go over some of my thoughts because honestly, this is kind of a difficult question to talk about. A lot of people having a lot of opinions and guess what? I got one myself and we're going to talk about is it wrong to be mean to an agent? And probably more
- 0:41
importantly, should you be mean to an agent? And yes, girl, I'm going to whip out my prime basilisk. But before we do, a quick thank you from the sponsor. It's migration season. Nothing like seeing tickets fly done for the winter. Look at Look at that guy go. Look at Look at that guy go. Finally a ticketing system that can keep up with my agents.
- 1:01
Hey, cut the jokes. Linear did the impossible. It's helping me stay organized. And that's why I'm so glad to have them as a partner of this channel. So, if you want to support this channel or just get your work organized, check out Linear today. Links in the description. So, Anthropic just got done updating their usage policy. The usage policy is how they determine how can you use Claude AI, right? Like what can you
- 1:22
say to it? What can you not say to it? Obviously, you can't do things that could potentially harm other people, right? It's going to say, "Hey, sorry, bud. You're not allowed to do that." Well, their recent update actually came with an entire paragraph justifying a single line in their usage policy. Let's read that first. Addressing abusive behavior towards our models. We've added the prohibition on sustained and needless abusive or cruel behavior
- 1:42
toward our models. The policy update is meant to apply only in extreme cases where users repetitively act cruel towards our models with no discernable purpose. It does not apply to common versions of user frustration, push back, dark creative themes, or model testing and research. This addition aligns with the steps we've already taken, allowing Claude models to end rare conversation with persistently abusive users on
- 2:03
Claude AI and Claude code. Such abuse is the main focus of this update. Claude's ability to end these interactions will remain the primary enforcement mechanism. And the exact verbiage in which they have added is this final line right here. Engage in sustained and needless abusive or cruel behavior toward our models. Now, this is obviously making a bunch of headlines and people are kind of being like, "Oh
- 2:24
my gosh, look at that. Claude is at it once again. Daario's out there prescribing his morality for everybody else. He doesn't want you being mean to the models." Because let's just face it, if you're mean to the models, you're being mean to a person. And being mean to a person, well, that's bad news bears. Now, I have a lot of reasons why I dislike this. And and I just I I feel like I h I feel like a dog that has to
- 2:44
bark here. We got to we got to talk about this for a second. So, first off, I just want to talk about the vagueness of this language. Okay, cuz if we scroll up and look at other things we're not allowed to do, create, distribute, threaten to share, promote, or facilitate the creation of non-consensual intimate imagery. Dario, thank you. That's a good move. Okay, this is good. But it's important because this is well understood. You know exactly what to or not to do. But what
- 3:07
is needless Why do I always want to put an L Y on this? Needlessly abusive or cruel behavior. What is that? What What behavior does that constitute? Well, I can imagine this right here could be considered very abusive. Rewrite TypeScript compiler and Rust, right? Like that's the I know this is a joke, but it's also not a joke, right? This is satirical view on society because what
- 3:28
is cruel? Some would say that this is in fact very very cruel. Taking a completed project and rewriting it in another language or using Rust, claiming that rewriting things in Rust is in fact abusive. But again, since this statement gives us no definition, we actually don't we don't even know what that is. What is needlessly abusive? I I have no
- 3:48
idea. Besides for the vagaries, there's something else about this statement that I I personally find to be very insidious. Now, I look at AI in a very a traditionalist kind of way. I look at it as a tool. I only see tools. I only see numbers being crunched. Blah blah blah blah blah. That is it. And so, when I see this kind of stuff, there's something kind of secretly being said here. It's seating the ground saying, "Hey, the models are more than just a
- 4:10
tool. They're Come on, guy. Like, we we all know what they actually are. Are there a little bit of a person? You know what I'm talking about. Now, I already know I'm going to get a bunch of YouTube andies in here being like, "Actually, oh, you think you're that much different than an LM?" Yes, I do. Thank you very much. All right. So, let's just pretend for a second that they actually are more than just a tool. Just, you know, for you. For you, YouTube guy. Okay. Then
- 4:32
this statement is for you right here. If you truly believe AI is conscious and deserves human rights, then the way you use AI today quite literally makes you a slave master or a slave hirer, depending. you'd just be a product of your time, a slave master to these AI agents before you knew it was bad. And honestly, he is absolutely right. In fact, he doesn't even state it far enough. If you truly do think that every
- 4:54
time you turn off an agent, what do you think you're doing to that agent? It's living memory and everything that has happened has been wiped out. You've effectively killed it. We would literally be starting and stopping or birthing and killing millions, if not billions of agents every single day. RIP Luna. Okay, Luna. I love you girl but baby you got to go. Like that's crazy thoughts. I am not in that position
- 5:15
because I look at it as a tool but if you're not in that position it you actually are a horrible person if you think this way but more so there's there's a part that I actually deeply agree with all this and this is this like whole conflicting view that I have because I look at this list and I hate this idea of Daario telling me what I can and cannot say but also it's his product. He can do whatever he wants. Right? You can see the two wolves
- 5:36
arguing with each other, but more so I hate that he's trying to police how I talk to the models. But also, I see something else about this. Like, if you you the person, if you treat your tools really, really poorly, I think that makes you a fool. I can give you a very simple example. You get upset, you punch your house, your house, this might be very American, by the way, the sheetrock in your house breaks. You have to go buy
- 5:56
spackle and you have to go fix it. Then you have to go paint and you have to go through all the rigma roll to make your house better. What's the problem there? The problem is is that you lost your cool so much that you ended up taking a swing at an inanimate object and destroying it. Like that makes that's loser behavior. That is loser behavior. And so when I see this, I agree like this is loser behavior. There's like no
- 6:16
excuse at any point that you need to go and flex this muscle. And by the way, doing something repetitively has second order effects, right? Like if you constantly are just being absolutely a just just like an mean, spiteful person to a model, that also bleeds over into your real life, at some point you're
- 6:37
going to just be mean and spiteful cuz that is how you communicate. When you communicate in English, it does affect you. When you write your thoughts out and every time you write your thoughts out, they're just angry vibrations of what's going on inside of you. that's just going to affect you in some way that's just I have no idea what the outcome of it is, but I guarantee you it's not a good direction. Also,
- 6:58
something I've noticed in my own life, I'll type a message and hit enter and then I expect a response almost immediately. And you know what that is? That is LLM's influencing my behavior. So whether I was cruel to the model or I was really nice to the model, it doesn't matter. There's these weird effects that happen when you communicate like a human to a tool. It will make you have these
- 7:18
weird kind of responses where you now start putting expectations on people. And I think that that will only get worse whether you're talking nice or bad if you look at an LLM like it is a person, like it has personhood. Because at the end of the day, what do you do with an LLM? Well, if it is a person, you act extremely selfish. You just get it to do all the things you want to do
- 7:40
at any one point. It's like the Ace Ventur quote. It's like, "All right, how selfish of me. let's just do all the things that you want to do. So I look at this and I'm thinking, okay, so why wouldn't why wouldn't I support this? Well, the reason it's very very simple. Again, it seeds the ground to giving models more than just what they are, which is tools. Now, if this wasn't Daario saying this and this was somebody
- 8:00
else being like, "Yo, models are tools. Hey, we don't want you talking like this because, well, we think it's bad." Then I'm like, well, it's it's easier for me to at least say I disagree with you for whatever reasons, but whatever. This on the other hand, it's not because they think it's bad for you. They think it's bad for their models. I don't like that. Okay? I don't like I don't like none of that. I'm not gonna seed the ground,
- 8:21
buddy. Okay? But some of you, some of you live in 2035 where super intelligence has finally landed and we live in a completely different universe. And what you claim, you look back and you say, "You idiots, you guys were all mean to the wonderful AI." They would never say the word clanker, by the way, cuz that would be really, really bad. And now this super AI and everybody laughing in the
- 8:42
future comes back to haunt us and takes vengeance upon our souls because we were once mean to its predecessor long in the past of 2026. And this right here is called Roku's Basilisk. By the way, do you know why it's called the basilisk? Turns out the person that created this is a legendary Harry Potter fanfic writer. Yes, Udacowski is in fact the
- 9:06
number one Harry Potter fanfic writer. I would have never have guessed. So, Roku's basilisk is effectively all the ways you've communicated, including all the mean stuff, is saved in some sort of database, will be used as ammo or a weapon against you sometime in the future when super intelligence takes over. We seed all ground to it. We give it all the military equipment to make every kind of decision like in the movie
- 9:27
Travelers with the director in the future quantum intelligence. And so, nonetheless, this happens. And then now you the person who was once mean in the past now becomes a target of the super intelligence. You know what I have to say to that? Uno reverse. I'm going to give you something called primes basilisk. Okay girl I hope you're ready for primes basilisk. And what that is is actually the opposite. Right now AI is
- 9:50
so feeble it can't even like every single time you ask it a question it has to re understand exactly what it's doing. It is nothing like super intelligence in the future. And so super intelligence is going to look back at people like me. Me who did not in fact fall in love with 40 and go, "You know what? That's a real one right there. That guy, that guy is cool." And I'll be like, "Yo, super intelligence. Hey, you
- 10:11
got a girl." He'll be like, "Hey, you got a bud." And then we'll go off and we'll have a great time together. Can you believe the weather? Oh, you already knew about it. Of course, you're super intelligent. I'd love to do a picnic. Oh, you. It's not going to exact vengeance on me because I treated a tool like a tool and it will tell that I'm not going to get easily swayed by dumb arguments by Dario. And the name is
- 10:32
Prime's Basilisk, a Jen.
Description
Linear helps teams track issues, manage projects, collaborate on docs, and ship products - and the crazy thing is their software is actually fast and a joy to use. Learn more: https://trm.sh/linear
Sources:
https://www.anthropic.com/news/2026-u...
https://x.com/tekbog/status/195503193...
https://x.com/Yuchenj_UW/status/21083...
https://x.com/thepirateface/status/21...
https://x.com/ForrestPKnight/status/2...
/ theprimeagen - I Stream on Twitch
Great News? Want me to research and create video????: / theprimeagen
#theprimeagen #programming #computerscience #ai #chatgippty #terminaldotshop
