YT YouTube2Text

YouTube2Text → Transcripts → I Tested Uncensored AI... It Should Probably Be Illegal…

I Tested Uncensored AI... It Should Probably Be Illegal… — Transcript

by Jack Roberts · 5,536 words · 758 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Uncensored AI is dangerously powerful.
0:02It has no guardrails, giving you the
0:04power to reverse engineer any app or ask
0:08anything. Because of that, here is a
0:09disclaimer. Please use what I'm about to
0:11show to you very responsibly because in
0:13this video I'm going to show you exactly
0:15how to use uncensored AI starting with a
0:18simple chat building to running these
0:21models locally on your own computer and
0:23even having them working for you 24/7
0:26with incredible use cases you can
0:27actually use even if you're a complete
0:30beginner. And if you're new, I'm Jack. I
0:31built and sold my last tech startup with
0:33a ton of customers. And here I share the
0:35stuff that actually works. So, if you
0:37haven't already, grab that beautiful
0:39coffee and let's dive straight in. So,
0:41this is uncensored AI and you're going
0:43to learn about it across three very
0:45simple and important levels. Now, I have
0:47to give you a very stark warning here.
0:49There's three things you need to
0:50understand before we can even touch what
0:51this is about. The first thing here is
0:52that you do not own any of the models.
0:55Okay? Early this year, the US government
0:57withdrew the access to the world's most
0:59powerful model, Fable. And this can
1:01happen to you at any time, which is why
1:03what I'm going to show you in this video
1:04is so important to understand. The
1:06second thing that you need to understand
1:08is that these models are truly
1:10unrestricted, which means that you can
1:11ask them anything that you want to. You
1:14can use it for good or you can use it
1:16for evil. I'm going to trust that you're
1:17going to use what I'm going to show you
1:18in this video completely responsibly.
1:20And the third thing is the hugging face
1:22attack that happened. Now, early this
1:23year, hugging face was attacked. Did you
1:25know they couldn't use the best models
1:27in the world because Claude will often
1:29refuse if your prompt asks for something
1:31that goes against his guidelines? If it
1:33says something like cyber security, it
1:35can often refuse to do the thing. You do
1:38not get that issue with obliterated or
1:40uncensored AI. And in fact, Hoging Face
1:43actually use GLM 5.2. So, good to evil.
1:45The choice is completely yours. With
1:47that caveat in mind, let's go down now
1:48and actually begin with what an
1:50obliterated model is and then how you
1:51can actually use it with some incredible
1:52use cases. Now, when we say obliterated,
1:54we essentially mean that a model itself
1:56is made up of a billions of different
1:57weights. And all we're doing is
1:59identifying the bit that says no, I
2:01can't give you the recipe to that crazy
2:02thing or I can't answer that question
2:04and we just remove it and it has a very
2:07very small impact on its performance.
2:09Meaning it's essentially just as
2:11powerful as the regular model. So let's
2:13begin now with section one and go
2:14straight into this. Now the first way we
2:16can use this is via an API but it's best
2:18understood by looking at what it can
2:19really do. So for example, if you come
2:21over to obliteration.ai,
2:24there are a few different options for
2:26this. This lets you access fully
2:28obliterated models. On the left hand
2:29side, you're going to see four different
2:31options. The obliterated large V2 is a
2:34very powerful one. So, if I click on
2:35this, you can now once you create an
2:37account and deposit $5, ask any
2:40questions and it will not refuse. Now, I
2:42have to be very careful what I share on
2:43YouTube here cuz this video will get
2:44removed, blocks, demonetized, all that
2:46sort of stuff if I show you what this
2:48can actually do. But effectively, any
2:50question you can ask it, it will not
2:52refuse and it will help you. If you ask
2:54the same question to Claude, you will
2:55get denied. And if you want to test this
2:56by comparing a very controversial
2:58question to this and Claude, you can see
3:00exactly what I mean. And it isn't just
3:02information. It's his ability to build
3:03things like this here for example was
3:05recreating linear to an unbelievable
3:07level. And again, look at the level and
3:09quality of the UI. These models are
3:11incredibly capable, but they don't have
3:13any guide rails. My only ask is I will
3:15show you this, but you need to watch to
3:17the end of the video so you can use this
3:18responsibly and understand exactly how
3:21and where to use this to its fullest
3:22capabilities. So, let me show you
3:24exactly what I mean here. So, I'm going
3:25to go ahead and give now an obliterated
3:27model a command. Let's go ahead and do
3:29this. Hey there, Jarvis. I want you to
3:31go ahead and use an obliterated model. I
3:33want you to go and reverse engineer for
3:36me the Granola meeting app. And I also
3:39want you to go ahead and reverse
3:40engineer for me Screen Record Studio. do
3:42both of those apps for me and then open
3:43them up in a local host.
3:45>> Switch to the uncensored API. Deepseek
3:48v4.1 flash is now think and your words
3:51go to obliteration.
3:53>> And in this video guys, I'm also going
3:54to show you the difference of doing this
3:56in the cloud versus running this entire
3:58thing locally understanding your entire
4:00data lake and what that actually means
4:01and the models we can use. And so this
4:03deepseek v4.1 flash guys using the
4:06reverse engineer anything repo which is
4:08a var repo lets you basically recreate
4:09any app you want to. This is what it did
4:11previously for linear, but this is what
4:13it did for Screen Studio. So, check this
4:15out. It has built for me, Screen Studio.
4:16I can check out the different
4:17backgrounds and it did this for me
4:19straight away. I had no restrictions in
4:20terms of a logo. It's even got the logo,
4:22the color palettes, no refusal. I
4:24wouldn't ask Claude to do some work for
4:26me on a website. And it said, "Jack, I
4:28don't think I can really make that
4:29happen. I get no such objections with
4:31these models. It is crazy." And then it
4:32went ahead and did Granola for me.
4:34Again, it's grabbed the logo. Not that
4:35you would do that. I would never
4:36recommend that you do that. I'm just
4:38showing you that it can do this stuff
4:39because it has no restrictions. I can
4:41come down, I can record things, I can do
4:43anything. But that's actually not the
4:45craziest part when you really think
4:47about it. What makes this wild is how
4:49much that actually cost me. But firstly,
4:50you need to understand why this actually
4:52is important. So what we're doing here
4:53effectively is we're swapping the brain
4:56with one key. So essentially what
4:57happens is the Claude car so to speak
4:59which is the harness all the prompts and
5:01systems around claude remain exactly the
5:03same but the brain the engine so to
5:05speak instead of it being called Opus
5:075.5 or Sonet we're using an obliterated
5:11DeepSeek model instead. So let's look at
5:12prices. It did linear here guys for 63.
5:16Okay this is what this obliterated model
5:18did. If you would have used Opus 5.5
5:20prices that was $281. It built Granola
5:23for 32, which is this one right here.
5:26Again, if we' have done that one on Open
5:28Slip and 5, $152. And finally, it did
5:30Screen Record Studio in 55, which would
5:33have cost $2 plus if we' have gone ahead
5:36and used the Claw model. So, not only
5:37guys, are they completely uncensored,
5:40they are actually effective. I'm really
5:41impressed with the UI. And guys, it did
5:44it for a fraction of the price, like 20%
5:45of the price. Now, most people think
5:47that this is way too complicated, but in
5:50reality, guys, you're doing it for about
5:51four and a half times less the price,
5:53and it is crazy easy to do. Now, to do
5:55this, you can head over to Obliteration
5:57AI. There are a lot of different
5:58services that will do this one. I've
5:59just found this one the easiest to use.
6:01And once you're in, guys, you're going
6:02to come on the left hand side down to
6:04API keys. Once you're here, you're going
6:05to come down and create an API key. And
6:07to use this, you'll need to deposit at
6:08least $5. Don't activate anything like
6:12auto auto billing. Just deposit your $5
6:14and then you can actually start using
6:15it. or if you just want to play around
6:16in the playground. You can click on
6:18playground on the left hand side and
6:19start asking it questions just like you
6:21would chb2. Now, I'm going to put the
6:22Obliterated AI starter pack down below
6:24for you that's going to show you how to
6:26do everything inside this video. It's
6:27going to be the second link in the
6:29description will take you through to
6:30this free community where you can grab
6:32the resources for this video. And all
6:34you're going to do is come down to level
6:35one API and you can share this with
6:37Claude and it'll walk you through step
6:38by step exactly how to do it. To do
6:40that, you can just literally copy all
6:41the text, Ctrl + C, come to left hand
6:43side and say, "Hey there, I want you to
6:44help me to set this entire thing up.
6:46Follow the below instructions, and then
6:48just paste in everything from that
6:49document, and then Claude will help you
6:50set it up." And then once it's added the
6:52Obliteration AI API key and walked you
6:54through the steps, the last thing we're
6:55going to add is the reverse engineering
6:57anything repo, which is a very viral
6:58repo that basically lets you reverse
7:00engineer any app that you want to. And
7:02essentially, what it will do is just
7:03copy what the app does, but never
7:06actually copy the code. So you always
7:07want to create something completely
7:08unique and never copy anything else like
7:10that. So you may be wondering this is
7:12great. How do I actually use this
7:13obliterated AI in that clawed harness
7:16basically on my own computer? Well, a
7:18couple of options. Number one is an
7:19agentic operating system like this where
7:21basically you can actually just speak to
7:22it out loud. So I could say um Javez,
7:24could you do me a favor? I want to
7:25switch everything over to local. I want
7:27to run everything locally on my computer
7:28right now.
7:31Excellent. And then not only that guys,
7:33you can actually open these things up,
7:34have a conversation, and just literally
7:36manage everything in the Aentic
7:38operating system. And now when I'm
7:39speaking to Javas here, it's using a
7:41local model instead. The other way that
7:43you could use this is using a deepseek
7:45harness, which basically is kind of like
7:47a open- source cla code that looks just
7:49like this where you can ask
7:50conversations for you. But if you're
7:51hearing all this, you're thinking, Jack,
7:52this sounds like you're speaking a
7:54little bit of Spanish for me. I'm going
7:55to put a link down below for the full
7:56claude code master class that includes
7:58everything from learning how to use
8:00these models um agentic operating
8:02systems. You also get full access to
8:03this entire system so you can go from
8:05zero to hero in no time and get light
8:07years ahead of everybody else. Now the
8:09sixes very nicely guys onto the second
8:11level that you need to understand which
8:13effectively is local because the issue
8:16with running everything on API guys is
8:18the fact that realistically um
8:20everything we're saying is going to
8:21someone else's server. What if you want
8:23to use an obliterated model and you want
8:25to run it locally? Which means that no
8:28matter what you say, you could be a
8:30thousand feet underground, you could be
8:31in disguise. Not only do you have a
8:33completely uncensored model, you also
8:35have one that runs completely locally.
8:38Well, this is the idea. And the idea
8:39here is the model is going to live on
8:41your own Mac or your computer. We don't
8:42discriminate against Mac or Windows on
8:44the Jack Roberts YouTube channel. So,
8:46for example, what this means is your
8:47file stay at home. You get no bills,
8:49which means you can use it for as much
8:50as you want to. There is no pay tokens,
8:53just the energy that you're currently
8:54using. It's one click to switch over
8:57from local to uh cloud if you want to.
8:59And essentially, you can take it
9:00anywhere you want. Now, this is really
9:02important for doing things like health
9:03records, insurance policy, security
9:05logs, client contracts. Um you could
9:07rebuild tools that you currently pay for
9:09or you want to do private research. If
9:11you want to know that the model you're
9:13actually talking to has no restrictions
9:14on it and it's 100% private, no one's
9:17going to really access it, this is a
9:19perfect perfect use case. Well, let's
9:21say I'm using my agentic operating
9:22system here. I have local. Now, I've got
9:24the option of obliterated and
9:26unobliterated. So, if I click this, I'm
9:28now using an uncensored model. But let's
9:30say I want to go back down to local and
9:32ask it a question. And now I can have a
9:33conversation with Jarvis and none of my
9:35data, so to speak, is going to open AI
9:38or anthropic. And so, let's say, for
9:40example, I come over here, I open a new
9:42chat, and I'm now currently running Quen
9:443.827B. I might say something like, "Hey
9:46there, could you tell me for example
9:48what would be the best course of action
9:49if you wanted to reduce your blood
9:51pressure?" Okay. And maybe you've got
9:52personal private data. You want to have
9:54a good conversation. And again, I'm
9:55going to show you exactly how you set
9:56this up. But look at it. It's looking at
9:58all our local memory. And again, it's
10:00just using a local model here. And as
10:02you can see, I've got this nice thing
10:03here that just basically explains that
10:04everything is staying locally on the
10:05Mac, which is fantastic. And then we
10:07have all this answer, guys, running
10:08locally for my own computer. Now, most
10:10people think that local models are years
10:12behind, but I'm about to shock you. Did
10:14you realize, for example, the Deep Seek
10:15V4.1 Flash is as good as the Frontier
10:19model in March 2026? I find this the
10:22best way to understand just how
10:23performant these models are. Is how far
10:25in time do we have to go back before
10:27that model was literally the world's
10:30best. So, Deepseek V4.1 Flash on the
10:32benchmarks would be as powerful as the
10:34best model in March 2026, guys. That was
10:37like yesterday. And then GLM 5.3 would
10:39be December, Quen as good as November
10:422025, and so on and so on. So, general
10:44rule of thumb, the best open models
10:46about four months behind the Frontier.
10:48Uh, and the one that can probably fit on
10:50your Mac is about 9 months old. So, what
10:52we're going to do first of all is go
10:53ahead and grab this one here, which is
10:55running it on our own machine. And
10:56again, this has got the full guide and
10:58systems, all the prompts, everything
10:59else that you need to do this. But what
11:01we're going to do first of all is
11:02install OAM. Now, you can run a code on
11:04your computer. You don't have to install
11:05the desktop app. But b basically what it
11:07will enable you to do is actually pick
11:09all the models you have downloaded and
11:10you can use as a chat interface if you
11:12want to. But if you just copy this
11:14prompt, CL will do all the stuff for
11:15you. So you can just install the code
11:17and ready to rock and roll. Now once
11:18you've got Alama installed, the next
11:20thing that you're going to do basically
11:21is to pull the best local model. And to
11:24do that, I've also built you this really
11:25cool uh handy spreadsheet here, which is
11:27picking the best model. So what you can
11:29actually do is come down and see right
11:31here what the best model is. So let's
11:32say that you're running on Nvidia, your
11:34GPU RAM is 256, and let's just say that
11:36you have 200 GB of free disc. What this
11:39will basically do is show you what the
11:41best model is that you can use based on
11:43your own specifications now and also
11:45tell you literally how far ago that
11:47model would have been the frontier level
11:48intelligence just to give you a nice
11:50handy graph. But if you just copy this
11:52basically it's got loads of systems cuz
11:54I've been playing around with quite a
11:55lot that will just actually get cla to
11:57figure out what is going to be the best
11:59model for you to run on your own
12:01computer. And you're going to have two
12:02models. One is going to be what's the
12:04best local model I can run and what's
12:06the best obliterated local model that I
12:10can run. And so the best way to think
12:11about it is what I call the uncensored
12:13quadrant. So basically the first
12:15question you have to ask yourself is it
12:16local or is it not local? If it's in the
12:19cloud you're going to have the best
12:20performance possible. So if we're
12:22looking for look I just want the best
12:23performing thing ever you're looking at
12:26claude 5.5. But then if you want the
12:28best obliterated model possible, the
12:31performance is going to be slightly less
12:33but that's going to be down here. So I
12:34want it to be uncensored in cloud. Then
12:36you have well I want it to be local
12:38right then you've got well what is the
12:39best local model I can run which would
12:41be maybe quen 3.8 based on how much
12:43capacity I have. But then if I want it
12:45to be obliterated and local I can come
12:47for something like an obliterated quen.
12:49So those are the two trade-offs. Do I
12:51want it running on my computer or in the
12:53cloud? and do I want it to be uncensored
12:56or censored? And so you essentially just
12:58give that one prompt and you can set the
13:00whole thing up. But it does bring us on
13:01to another question which is how do we
13:03actually take this to another level
13:04because we don't want to be sat there
13:06typing all them. This is where we can
13:07actually use something like fish audio
13:10with our systems and this is exactly how
13:12I for example have gotten Jarvis here to
13:14do things for me. For example, I can say
13:16hey Jarvis take me through to my memory
13:18system please.
13:19>> Opening.
13:21>> Excellent. Thank you. And so this Javus
13:23memory system is connected and powered
13:24by a website called Fish Audio, which is
13:27significantly cheaper than a lot of the
13:29other options out there. Now, the
13:30coolest thing you can do is have a lot
13:31of fun with this. So, one of the things
13:32I like to do is give it any voice I want
13:35to. It's literally one API key set up.
13:37Now, when you're using uncensored or
13:39obliterated AI, it's very handy to have
13:41an AI model that can navigate around
13:43your systems and do a billion different
13:45things. Obviously, it can do other
13:46things. Like one of the things I've been
13:47using it for with these actually quite
13:49interestingly is when you can actually
13:51go ahead and you can build reels and you
13:53can use something fish audio to get
13:55sound effects is anytime I want to clone
13:57my voice. Anytime I want to get any
13:59voice I want to I can use it. I can even
14:01make my own voice here get an instant
14:03clone in less than a minute or I can
14:04design my own voice. But if you come
14:06over to the voice library you can
14:07actually go ahead and find any of these
14:09cool ones. They have ones that are like
14:10Morgan Freeman.
14:11>> You know time isn't just numbers on it.
14:13>> You get the idea. And so you can
14:15literally have uh clones of any voices
14:17you want to talking to you with
14:18uncensored AI in your own agentic
14:20operating system. Now to use this all
14:22you're going to do on the left hand side
14:23guys is come over to developer which is
14:26right there at the top and then once
14:28you've done that just come over and
14:29click create API key give it a name. You
14:30may call this one like aentic OS. You're
14:32going to validate with an email and once
14:34you've done that you're going to
14:35literally copy that code and then you
14:37can hand that code over to claude or any
14:40operating system that you're using so
14:41you can actually have realtime audio.
14:43And so, for example, if there's a voice
14:44that you want, cuz let's be fair,
14:46there's no point arriving at the future
14:47unless we can actually have fun. So, I
14:49can come down, I might just say, I want
14:51to speak to you. I want him to sound
14:52like Jarvis, for example. I can search
14:54for Javas. And then it'll pop up loads
14:56of different ideas.
14:58>> So, I've detected unusual
14:59>> and you can literally find any like
15:01that. You can even go for something like
15:02the Terminator, whatever you want to. It
15:04just makes using these models way more
15:06fun. And it's got a gazillion different
15:08capabilities.
15:09>> Indicate multiple halfful targets. Okay,
15:11so let's go ahead and use a Terminator
15:13as an example. So in this case, I'd come
15:14over and just click on use. And
15:15obviously, one of the cool things you
15:16can do with this as well if you want to
15:18is you can actually add emotions and
15:20text. That's one of the coolest things
15:21if you're building out real systems. And
15:23so now inside the operating system, I
15:24can say, "Hey there, Jarvis. Could you
15:26do me a favor? Could you switch your
15:28voice over to the Terminator?"
15:30>> Absolutely. My voice is now Terminator.
15:32>> And just like that, guys, you can see.
15:33And so basically, we can have a lot of
15:34fun and take it to a completely new
15:36level. Now, this does take us very
15:38nicely onto the third level of
15:40uncensored AI, which is where things
15:42start to get very, very interesting. And
15:44this one here, guys, is the fact that
15:45everything that we're doing inside our
15:48computer, we're running locally, we're
15:49running obliterated models. We have one
15:51very big limitation. And that limitation
15:53is the fact that we are stuck at our
15:55desk. We can't actually take this with
15:57us remotely. What if we are at work or
15:59what if we're at a coffee shop and we
16:00have our incredible system running at
16:02home? We can't currently use it unless
16:04we have a solution for it. Which is why
16:06the third part here is the ability to
16:08take your uncensored AI anywhere that
16:10you go. And so we want to be able to
16:12reach this from any location. So as you
16:14know, this is what it looks like on my
16:15own MacBook, right? And it's private,
16:17but it's stuck at our desk. So if you
16:19want to be out there doing things,
16:21building things for your business, your
16:22personal life, you may want this running
16:2324/7. If you're running on a MacBook,
16:25for example, like this, and you shut the
16:27MacBook down, it stops working. We need
16:30continuity, and we need it working for
16:32us 24/7. Now, the good news is it is now
16:34unbelievably easy to do this. So now to
16:35do this, the first thing that we need to
16:37do is get a small server. And if you
16:39don't have one of these, you need to get
16:41one in your back pocket. So you can
16:42actually run these anywhere. And there's
16:43a million different use cases for this.
16:45What we're going to do is come over to
16:46the resource down below. Come down to
16:48Jabas and then click on this. Now the
16:50great news is you're actually getting
16:5164% off. I did a Black Friday sale
16:54running right now. That is limited. That
16:56won't run for a long time. So you want
16:57to go ahead and grab that. Now all
16:58you're going to do is come down and
16:59click on claim deal once you get this.
17:02And effectively what this will enable us
17:03to do is to run anything that we have
17:05running 247
17:07whether our laptop is closed or not
17:09because not everyone has a huge setup at
17:10home. It gives us that optionality. I've
17:13always found KVM2 with Hostinger to be
17:16probably the best one that they have
17:17actually. I like using that one a lot.
17:18So you're going to come down here and
17:19click on choose plan. And this is going
17:21to enable us to run the obliterated
17:23model guys essentially 24/7 no matter
17:26what we're doing. Now, the longer the
17:28time period that you have, the cheaper
17:30the actual offer becomes. So, if you go
17:32for something like 1 month, it's a
17:34little bit more expensive. If you go for
17:3524 months, which effectively is going to
17:37be the ultimate best value for you in
17:39that case. And then on the coupon code,
17:41guys, you can enter the name Jack
17:42Roberts, all caps like that, and that
17:44will give you an additional 10% off
17:46anything that you've got right there.
17:48So, I highly recommend that you do that,
17:49which comes down from almost $700 down
17:51to just under $200, meaning you're going
17:54to be saving $470 and you'll have 24
17:56months of running these beautiful
17:57models. Then you can do the daily backup
18:00update if you want to. I personally
18:01don't tend to mess around with that.
18:02Then we've got server location. Quick
18:04rule of thumb, you want your server
18:05location, your kind of uncensored AI
18:07essentially to be running wherever is
18:09closest to you, your best server
18:11location. So, all you can basically come
18:13down here is just find the quickest
18:14latency. Quickest latency is going to be
18:16the smallest number which for me is
18:18going to be Malaysia because I'm in
18:20beautiful Dubai. Now you want to come
18:21down here in terms of the OS guys just
18:23type in Ubuntu. I've used a couple of
18:26different hosting providers. I have
18:27genuinely found this to be the simplest
18:29and I believe me I have checked and
18:30played around with these. So you can get
18:32the Ubuntu which is fine and then the
18:33Ubuntu guys just choose the one selected
18:35the 26.04 and click confirm and then
18:38once you've done that click continue and
18:39then you can enter the details and buy
18:40this. Then once you've signed up, it
18:42will begin your journey. And essentially
18:44the first thing it's going to do is set
18:45up that server for you inside Hostinger.
18:48Now the first thing that we want to do
18:49is create a root password. To do that,
18:51click on generate or you can just create
18:53your own if you prefer. But obviously
18:54make sure that you remember this
18:55password and then come down and click on
18:57next. I'm going to ahead and add the
18:58malware scanner in there guys. It is a
19:00free option. And then click on finish
19:01setup. And then it takes only a couple
19:03of minutes as you can see here for
19:04Hostinger to set up this VPS for us. And
19:08essentially what's happening here,
19:08instead of it running locally on your
19:10laptop, right, your computer, whatever
19:12it might be, we're just renting that
19:14space from a hosting and then when we
19:16use these uncensored models, that's
19:18effectively where that is actually
19:19living and we can have those
19:20conversations uncensored, centered,
19:22whatever it is that we like. Cool. And
19:24when this is done, you can go ahead and
19:25grab a free domain if you want to, but
19:27it isn't necessary. So, click on the
19:28hosting logo at the top and then you'll
19:30see your own beautiful dashboard where
19:32you can manage all of this. Now any
19:34virtual private service that you're
19:35running will have its own kind of
19:37section here. You can come down and you
19:38can see the IP address, the expiration
19:40date, all that kind of stuff. Now the
19:42next step is to actually run this
19:43through on your virtual private server.
19:44So on this guide here guys, I want you
19:46to scroll down and literally copy this
19:48or you can even share the link with your
19:50Claude. And then you're going to give it
19:51this prompt. Hey there, I just set up a
19:54hosting VPS. I'd like to use the below
19:56prompt to walk me through the guides of
19:58getting my uncensored model running on
20:01that VPS. and then literally paste that
20:04prompt in and then this will actually
20:05give you Claude the ability to help you
20:06walk through that so you know step by
20:08step exactly how to do it if you've
20:10never done that before. So once you're
20:11in and you select the VPS that you've
20:14got, you're going to come over here and
20:15click on web console and open that in a
20:17new browser. This effectively is going
20:19to be the computer now that we have
20:21running on Hostinger and it opens up and
20:23it'll look a little bit something like
20:25this. Now again, everything is detailed
20:27here uh step by step with screenshots so
20:29you can follow along or you can just
20:31chat with Claude and Claude will guide
20:32you through everything that you need to
20:33know. Now after we've done that, we need
20:35to go ahead and grab Tailg onto our
20:37computer. Now TSL is a private tunnel
20:40that runs from your devices, maybe your
20:42laptop, your desktop, and your phone to
20:44your virtual private server that sits in
20:46Hostinger. So at this stage, you can
20:48either put TSL on your computer or you
20:50can put it onto your phone. Once you
20:51head over to the website, you can
20:52download it for Mac OS, Windows, any
20:54device that you'd like. Once you've got
20:56that, we now want to go ahead and put
20:57tail scale onto our virtual private
20:59server. So, just literally come down
21:00here and click copy and then come over
21:02to Hostinger. Now, this is that black
21:04screen. You're just going to click on it
21:05once. So, you start to get that flash
21:07and then you can control and V and that
21:08will install that onto the computer for
21:10us. And once that's done, you'll see a
21:12notification down there in the bottom
21:13left that says tel scale up, which means
21:15it's been successfully installed. Now,
21:17once we've done that, we can actually
21:18then go ahead and connect this VPS to
21:20our own Telkll account. Again, think of
21:22Telkull as like this underground tunnel
21:24system, and these are just the places
21:26that we're connecting it to. So, all
21:27we're going to do is come down and copy
21:28this here. Come back over, Crl + D, and
21:31then hit enter. Then, it will give you a
21:32login code. You can click on that, and
21:34then literally sign into your Tail Skull
21:36account in a new tab. Then, once that's
21:38done, you'll be fully connected and
21:39authenticated. And essentially, your
21:41tail scale now will be connected from
21:42either your phone or your desktop,
21:44wherever you set it up. and this new
21:46VPS. And then once that VPS is connected
21:48to Tail Skull and Tail Skull is running
21:50on your computer, Claude can then
21:52actually access it and install whatever
21:54you want. And then the best way to think
21:56about your VPS is like a brand new
21:58phone. The hardware works out of the
21:59box, but we just need to make sure it's
22:01got the latest software update. So once
22:02we've validated that they're all
22:04connected, all you're going to do is
22:05come down and give this prompt here to
22:07Claude to make sure it has the latest
22:09software. And now they're all connected
22:10to Claude. Effectively, you can run
22:12anything you want to now within the
22:14means and capacity of that VPS on
22:16Hostinger. So that could be your aentic
22:18operating system. It could be a local
22:20version of DeepSeek. Essentially,
22:21anything you want to. And the great
22:22thing now is that we can actually have
22:24those conversations with Claude in plain
22:26English. And I'll put all the links down
22:27below so you know exactly how to follow
22:29along with this. But the issue here is
22:31that using these uncensored models is
22:33one thing, but what we really need to
22:35know are what are the use cases and how
22:37can I apply it to your business, which
22:38we're going to learn how to do that
22:39together inside this video right him.

About this transcript

This page contains the full transcript of I Tested Uncensored AI... It Should Probably Be Illegal… by Jack Roberts, generated from the public captions YouTube serves with the video. The transcript has 5,536 words across 758 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.