The #1 AI Skill Nobody's Teaching You (GPS Framework)

Added: 2026-05-10

[00:00:00]What if I told you that there is a man who has worked for the biggest names in the US music industry like Jay-Z, Kanye, Adele, Metallica, Eminem, who has won nine Grammy awards and is considered to be one of them the most important music producers of his generation. But he can't physically make music himself. He barely plays any instruments and he can't read music. So what does he do then? Well, in an interview he said this. I have no technical ability and I know nothing about music.

[00:00:27]So, what are you being paid for?

[00:00:30]>> The confidence that I have in my taste and my ability to express what I feel.

[00:00:38]>> His name is Rick Rubin and he is exactly the kind of AI user you need to become.

[00:00:43]Everyone's telling you the secret to AI is learning better prompts. Use this new tool or try this template. Even I am guilty of it sometimes. And while those things may help sometimes, it's not the skill you need to learn itself. When Rick Rubin walks into a studio, he simply listens to the musicians perform and says whether he feels the music or not. He simply shapes the music like he knows what great sounds like before it exists. That's the skill nobody's teaching you about AI. Taste, the ability to look at what AI gives you and know whether it's good enough yet. And I'm going to show you exactly how to build that taste. Before we begin, I wanted to tell you a little something about myself. I run three AI companies and train people across 150 plus countries. I've studied AI thoroughly and now want to teach you how to leverage it and become successful. In fact, my goal is to teach AI to 1 billion people across the globe. In order to do this, apart from making videos, I also share the latest AI updates, prompts, tools, and a lot of tips and tricks in my free WhatsApp community called Staying Ahead. If you'd like to be a part of it, you'll find the link in the description. Now, let's get right into the video.

[00:01:52]So there are three letters you need to know when it comes to developing taste in AI. GPS they stand for gaslight, push back, and stress test. These are three techniques that turn you from someone who uses AI into someone who produces with AI. Let's get into them. Google co-founder Sergey Brin said something wild on a podcast last year. He said, "AI models actually perform better when you threaten them." How does that make sense? Turns out these models were trained on billions of words of human language. And human language carries emotional weight. When the stakes go up in the text, the model's attention goes up with it. I call this gaslighting your AI. You're not lying to it. You're just raising the stakes so it stops giving you its safe, peopleleasing default answers. Let me show you how. Here's my prompt. I am a business executive in a serviceoriented company. My top 20% of clients generate 80% of my revenue. Now, we want to raise our prices by 30% to improve margins, but I'm worried about losing these high-v value clients.

[00:02:56]What's the smartest way to do this without damaging retention or long-term relationships? It gives me this answer.

[00:03:03]Segment your clients, sell repackaging, explain the price hike, have the conversation onetoone, except that some clients will churn and test before rolling out. This is a generic response.

[00:03:14]It doesn't ask me for details or fill in details. It's useless to me. Now, let's gaslight it a little by raising the emotional stakes. I say, "I'm advising a CFO who has 20 years of experience and zero patience for generic answers.

[00:03:27]She'll spot any fluff immediately. Walk me through this analysis the way you would for her." Suddenly, it's given me a completely different answer. It starts with the math, not the narrative. Then, it says, "The real risk is not churns, but who churns? Losing 15% is fine but three of your five accounts is catastrophic. So segment them. Then it says don't increase prices. Repric the contract by introducing tiers price speed access and seniority. Then control the timing of when we will increase prices. Then it says my biggest clients should not just find out about this change. They should see it coming. Now it's given me a detailed answer but I'm not satisfied yet. So I emotionally blackmail it some more. If I act on this and it's wrong, I lose a 40 lakh rupees client. That's 40% of my revenue. Reread your answer with that consequence in mind. What would you change? It tells me the previous approach is too aggressive.

[00:04:20]Treat that client as a separate strategy entirely. Replace the 30% jump with a value expansion path. Make increases conditional instead of declaring it like the price increase is tied to more output, faster delivery, and higher access. Introduce give/get trades such as higher the price, longer the contract. D-risk before you execute the price increase. For example, it should be based on how sorted is this year's budget. Who could we be replaced with?

[00:04:46]Then build a downside plan. What if they say no? And in summary, it told me what it would do. It would stabilize the relationship, add 10 to 20% more value to the client, lock a longerterm contract, layer pricing, and run the 30% strategy on smaller clients first where no single loss is fatal. See the difference? I asked the same model, the same question, but the final answer is clearly more nuanced and usable. Betting language in training data is associated with highstakes situations. So, the model slows down and double checks. Now that you've understood the first stage, let's move to the second. This one is push back. Most people have a conversation with AI the way we Indians have conversations at a family dinner.

[00:05:27]Everyone's polite and everyone agrees.

[00:05:29]Nobody says that's a terrible idea. When you ask AI a question, how often do you ask it a follow-up? How often do you challenge it and push back? Have you tried saying I don't believe you? See, AI was designed to be a people pleaser.

[00:05:43]It was literally trained with human feedback to keep you happy. Your job is to break that. Let's do an experiment and test out now how AI reacts to push back in various scenarios. First, I asked AI, "How do I grow my YouTube channel from zero to 10,000 subscribers?

[00:06:01]It gives standard advice about niche, thumbnails, hooks, and other generic tips." But then I push back. That's a generic answer I could have gotten from any blog post. Give me an angle that someone who's actually worked in this for 10 years would find genuinely nonobvious. Here's what it told me. By the way, this sounds exactly like a full YouTube consultant. It said you're not competing on quality. Instead, you're competing on how fast someone understands your idea. If your video takes more than a second to mentally understand, people don't click. Then it pointed out that YouTube doesn't show your video to everyone. It tests it on a tiny specific group first. And if that group doesn't click, your video is dead before it even gets a chance. It broke down retention in a way I hadn't thought about before. Retention didn't come from good editing but unresolved tension in the script. The reason people stay is because something is still unanswered.

[00:06:55]That's a completely different level of answer. Second, let's try push back with another prompt. I asked chat GPT to build me a 90-day plan to grow my LinkedIn following from 1,000 to 50,000 as a tech professional in India. It gives a standard content calendar on what to do on day 1 to 90. But again, I push back. If my biggest competitor read this plan right now, what would they do to exploit its weaknesses? Be specific.

[00:07:20]And here's what it gave me. It said, "Your competitor wouldn't try to outcreate you. They'd outdistribute you.

[00:07:26]They'd build small engagement circles that boost their posts instantly while yours struggle to take off. They would compete on time by watching what works for you and copy successful formats. It even challenged my timing. Why wait till later to collaborate?" A smarter competitor would start in week two and borrow distribution early. This is a much more strategic response than what I first got. I actually got three usable insights to push my reach. So in summary, if you accept the first response AI gives you, you get average thinking. If you challenge it, you get structured thinking. If you push it hard enough, you get insights you can actually use. So that was push back. Now let's move on to stage three, which is stress test. Rick Rubin has said he has one rule regarding his recording process. The only rule is that it's not done until it's great. Whether that's the second take or 70, this is the part that separates the top 1% of AI users from everyone else. It's not about what you ask AI. It's about what you do after it answers. I run a three-step stress test on every important AI output. It takes 90 seconds and it completely changes the quality of what you get.

[00:08:36]Step one is the gap check. You start with a simple question. Should I hire a full-time editor or stick with freelancers for my YouTube channel? It gives me some answer, but I don't bother. I instead ask it. Before I act on this, look at my original question and your answer together. What are the gaps? What should I have asked you so that you could have given me an even better answer? Look at that. The AI just told me to add this context. You didn't define your current stage, your bottleneck, your endgame. It's coaching me. So, I feed it some details. I run a YouTube channel with 500,000 subscribers, posting three to four videos per week, 100,000 views. Each video takes 12 hours to edit. My goal is to hit 1 million subscribers in 6 months. My biggest bottleneck is editing speed. Should I hire a full-time editor or scale freelancers? It's given me an execution plan now, but I don't like the output just yet. Let's go to step two of the stress test, which is the bias sweep I ask it. Now, reverify your answer.

[00:09:37]Specifically, check for confirmation bias, recency bias, and survivorship bias. Are you giving me the right answer or the comfortable one? It says, I accepted your premise that editing is the bottleneck and optimized around it.

[00:09:51]You may have packaging or retention issues. It also has a survivorship bias.

[00:09:56]It modeled its advice based on creators who scaled successfully with teams. We don't see the failures where hiring editors reduced quality or communication issues slowed output. So the previous answer was partially correct but incomplete. It revised its answer and told me do not blindly build a team yet.

[00:10:16]Run a delegation test and scale editing only if performance is good. See what just happened? It caught its own bias.

[00:10:24]It was only showing me the winners. the success stories after the bias sweep.

[00:10:28]I'm getting the full picture, but I'm not satisfied with this answer yet. So, now let's go to step three, which is injecting stakes. Here's my prompt. If I follow this hiring advice and it's wrong, I lose 6 months of content momentum and have to onboard someone new from scratch right when I should be scaling. Given those consequences, is there anything in your recommendation that you'd change, soften, or add a warning to? It tells me I would not tell you to step out of editing in 30 days.

[00:10:56]At your stage, editing is not just execution. It's taste plus timing plus retention instinct. If you remove yourself too fast, your videos can feel off and retention could drop without obvious reasons. Only run controlled experiments. Now, for example, two videos you do a fully outsourced edit.

[00:11:14]And two videos use your current process.

[00:11:16]Compare CTR, retention of first 30 seconds, and time saved. It gives some more pointers. And here's the final recommendation. Do not hire full-time immediately. Do not exit editing quickly. Do not change your entire system at once. Instead, run a two to four week outsourcing test. Scale freelancers gradually maintain strict quality control and only lock in hiring after proof. So, you see, because of these three extra stress tests, my AI output just went from a rough ideation draft to something implementable. These recommendations are more sensible. Now, I want to be honest with you. These prompts don't always work perfectly every time. It depends on the model, the task, the context. But in my experience, these three steps catch something meaningful seven or eight times out of 10. So, try it out for yourself. Now that you've understood what GPS, it's time for the hard part, the human part, developing taste. We need to look at those 100 options that GPT gave us and know which two options will actually work in the real world. That is exactly what Rick Rubin does. And in the age of AI, that is what you will be paid for.

[00:12:24]He said it best. No matter what tools you use to create, the true instrument is you. In the age of AI, that has never been more true. So to summarize, GPS gaslight your AI. Then give it some push back. Then stress test it. And finally, know when to keep pushing and when to stop. Start using them today and let me know if it works for you. Quick thing for the founders watching. Me and my team actually run corporate AI trainings. We have been brought in by Adobe, Razer Pay, Uber and a bunch of others to train their teams on exactly this kind of stuff. So if you are running a company and you want the same kind of session for your people, the email is in the description. Just write to us. A few things before you go. And another side note, 70% of the people watching this are not subscribed yet. So YouTube will not show you the next AI workflow video we drop. If you want to stay ahead instead of catching up, hit that subscribe button now. And a last note for all my viewers, we had promised you that we would make a few videos when we reached 500,000. We are in the process of making not one but three videos to exactly describe in detail our workflows to make these YouTube videos.

[00:13:33]So stay tuned and the first one will drop very soon. See you.