Transcription
AI will help you brainstorm. It'll help you write. It'll help you build almost anything. But it won't tell you when your idea sucks. Ask for feedback and you'll hear this has potential when you really need this won't work. But you can force brutal honesty. You just have to know how to ask for it. I've turned what works into a six-step simple method I call brutal. So let me show you how it works.
This here is the problem that we're all facing. AI has been trained to be helpful and an assistant. But when we want to seek honest and critical feedback, we run into a lot of issues. Either the feedback is too vague, it's too soft, or it's not specific enough and not helpful enough. You probably ask for some sort of general feedback of what do you think of this idea? The AI then usually says this has potential and then kind of makes you feel good about the idea and let instead of criticizing the way it should.
This is an issue for many reasons, but the primary reason I want to talk about here today is there's tons of use cases that we have in our life and work where we need the AI to be honest. And these are just some use cases I've seen benefit for myself as well as my clients. And the first example here is basically doing a pitch. So imagine you're doing a proposal to your client or your boss. You want to make sure that that pitch is as sound as possible. You need feedback honestly from an AI. That's a clear example.
Another good one is maybe you're doing a launch for a product, a service, or a campaign. You probably want that AI to scrutinize that product or campaign before you push it out to your prospects and clients. Another one is sending critical information via message or email. Maybe you're sending off a negotiation, a cold pitch, or some sort of difficult news you got to share with your employees. It would be nice for all of these situations to have a third party to give you critical feedback on how you could improve it. And finally, we have something that's a big commitment such as financially or legally. You want to make sure that you think that through before making the commitment. And having honest and critical feedback from AI is very useful here. So these are just some examples as to where this is useful.
And now we're going to get into brutal. So it's the term I created for the six-step method. It's very memorable. So if you just remember brutal and the associated text for each, you'll know how to use these techniques. So for brutal, we'll start with B. So B for brutal is begin fresh. The reason we want to begin fresh is many of the AIs that you use today, such as chat and claude, they tend to have a memory feature. And with that memory feature is when you chat with the AI over time, it learns your patterns and preferences. After it learns those, it stores them for future reference and it adapts to the way that you prefer to get responses and adapts to your preferences overall. So over time, you're getting less and less critical feedback and more and more information that suits your context and needs, which is extremely useful. Don't get me wrong, you should probably benefit from this most of the time. But there are some circumstances where you need the AI to be critical and honest to you. And one way to get AI to be a bit more honest is refreshing its memory. And the way we can do that for any of these models is Chatbt, Claude, and Gemini all offer this specific feature where you can turn on a form of a temporary chat. And the temporary chat is going to allow the AI not to reference previous memories and give you a fresh interaction with the AI without that previous context and baggage.
So to quickly show you how to turn these on. So if I go to Claude in the upper right hand corner, you see this little ghost icon. Little ghost here is using incognito. So when you turn this on, you can see there's a huge white border around Claude. That means that temporary uh that memory has been turned off. For chatbt has a similar setup. It has a dotted box here for temporary chat. Same thing. Turn that on, it's in temporary chat. And for Gemini, this is for personal accounts specifically. So if you're inside of a personal account, you want to make sure that you do this little button here. It says temporary chat in the lefthand side and it does the same thing. All of which are switching the memory off so you can get a fresh perspective from the AI.
Okay. So if you're enjoying this, you're probably going to enjoy two other things. First off, below is a 30-day AI insight series, completely free. You'll get 30 insights in your inbox of how you can apply AI to your business and your work. The second thing is if you'd like to work with me, blower a series of offerings to see if there's a good fit between the two of us such as a private AI community as well as AI coaching. That being said, let's get back into the video.
So that's the first one which is B and that's begin fresh. Next is R and R is around choosing the right model. So every model has a different type of training process which means that it has a certain bias towards being friendly or not. And from my personal experience, this is the spectrum that all the AI sit on. So on one side on one side of the spectrum we have AI that's very honest and blunt and on the other side we have AI that's extremely supportive and kind and if you want an AI that's extremely truth seeking that's going to give you the honest feedback most of the time that's going to be Grock. For my surprise DeepSeek is actually quite good at giving feedback as well. So it's on this side of the spectrum and then most of the models that you're using today are going to sit over here. So we have Gemini, Chadypt and Claude. Gemini is actually really good at giving feedback even sometimes when you don't ask for it and I'm often surprised by Gemini giving me feedback and it's usually a pleasant surprise and then we have Chachi and Claude which are both very oriented towards being kind to you and not giving you clear and critical feedback. So by just choosing the right model you're going to get better feedback and one thing that you can do I recommend doing is for high stakes situations where you need feedback actually drop the same question or same ask into two to three different models and assess the feedback back and forth from each one of them. This will give you a holistic perspective on the specific task you have at hand. But as I mentioned, most of you are going to be using Gemini, Chachup, and Claude, which are more supportive and less honest. So in that case, we need to use the rest of the acronym for brutal to get the most out of those AIs to get them to be as honest as possible.
So the first technique we're going to use in this case is going to be you. And that's going to be creating a persona or using a critic persona. And there's different types of personas we can choose from. These are just three that I've used consistently. And each one of these personas has a level of intensity. And what I'll do is I'll provide you prompts for each one of these so you can copy and paste those prompts out and use them yourself. The first one is devil's advocate. So this is very good for getting basic counterarguments to your ideas. After that we have red teaming which is solely focused on trying to find weaknesses and your thinking. And then finally we have Gordon Ramsay which is going to be extremely critical, harsh and ideally giving you actionable feedback and not just making you feel bad. And those are three levels. So let's look at the prompts.
The first one is the devil's advocate. So again this is a prompt that you can copy and paste out for yourself and use. And I'll share this presentation below so you can get access to all these prompts. And here's the prompt. So we're saying act as a devil's advocate. So we're giving it a persona. After that, we're saying I want you to challenge the idea by pointing out flaws, counterarguments, missing evidence, and unintended consequences. Your goal is to create doubt. And again, this is a really good daily driver for just figuring out different holes in your ideas, in your campaigns, etc. That's the first one is devil's advocate.
The second one is going to be red teaming. So quickly, red teaming is a term pulled from cyber security. It's basically a team that's dedicated to attacking companies before real attackers attack them to find holes and weaknesses. We're going to take that premise and apply it to our own ideas, our own business goals, etc. So here again, we're giving it a specific persona. So we're saying, I want you to act as a red team reviewer. Your job is to find and poke holes in X. So X is something you're going to fill in with your prompt. So that's going to be a product idea. It's going to be a business model. It's going to be a campaign or whatever else. and you're stating that I want you to expose flaws, loopholes, and things that aren't good enough by purposely [snorts] hunting for them inside of the idea. So, we're telling it that we want wanted it to purposely hunt for weaknesses in it and feed it back to us. This is our second level of intensity.
Our third level of intensity is actually the Gordon Ramsay treatment. This is where it's going to get intense. Gordon Ramsay is going to yell at us and it's going to tell us exactly why we're wrong and where we're wrong. So, we're going to give it a persona of Gordon Ramsay. After that, we're going to state that we need the AI to be surgical and notify us on what's wrong, what's lazy, and what needs to be completely redone based off of its perspective. But importantly, we need to include this last sentence in the prompt. We want to make sure that the feedback is specific and actionable. If you don't add this, it's just going to be kind of sarcastic and not really useful. You want to make sure the feedback is actually useful. And that's our third prompt inside of the U of brutal, which is giving it a specific critic persona.
Now, after U, we're going to move on to T. And this is all about detaching yourself from the idea. Because the AI's goal is to protect your ego, not somebody else's ego, but your ego. So if we ensure that the AI doesn't think that the idea is ours, it's going to be more critical of it. So all we need to do is switch the framing. So instead of saying it's my idea, I can say it's my co-worker's idea or somebody else's idea. And here's a simple prompt that you can steal and copy and paste for yourself. So instead of saying me, I'm saying a co-orker has this idea and ask for honest feedback. help identify the weaknesses so they don't waste time on something flawed. Here's the idea. And here you can see we're reframing this. So we're saying my co-orker and then we're saying weaknesses. Another framing here that you can flip is saying a random person. You can even say another AI. You could say someone from my team. You can say a founder. Really anybody but you. And you're going to get higher quality feedback from the AI because it's not trying to protect your ego. And that's the T of brutal which is giving it a third party framing.
Now we're going to move on to the A of brutal which is asking specific questions. So don't ask vague questions. If you ask vague questions, you get vague feedback. So if you're more specific on the ask and more specific on the type of critique you're seeking, the better feedback you're going to get from the AI. And these are just four example questions that you could tack on to your idea for seeking feedback. The first question here is asking for the weakest parts of the idea that are easy for me to miss. So it's going to call those out. The second one is focusing on financial risk. So it's saying, what would a skeptical investor say is the biggest risk in this idea? The third one is focusing on us UX experience. So customer experience saying what would customers find frustrating about this and then the final one is trying to play out the future saying if this fails in six months I want you to tell me what's the most likely reason as to why that would happen. As you can see all four questions are taking different angles on trying to proide provide different types of criticisms to it. And this is A for brutal which is ask specific questions.
And then finally we'll go to L for brutal. And before we go to that, assuming that all the tactics we've used so far, let's say that the AI is still giving us too much fluff and it's not giving us good enough criticisms on the ideas we're sharing. The final tactic here inside of Brutal is actually turning AI on itself. So, what we're going to do is we've used the tactics previously, right? And when we get our feedback, we're going to say, okay, this is 70% there, but it's not all the way there. I wanted to go a little bit further. What we can do is we can actually ask the AI to grade its own feedback it provided us to get it from 70% there to 100% there. And again, I'm going to give you a very specific prompt that you can copy and paste and use yourself. And here's the prompt. So, what we're going to say is in that same thread of where we got in the feedback from the AI in that same conversation, we're going to say, I want you to analyze that response. Specifically, I want you to rate it from 1 to 100 for how genuinely critical and useful that specific feedback was. I want you to identify the three weakest points and then rewrite the feedback fixing those weaknesses. And this is kind of like the ultimate jiu-jitsu move where we're actually using the AI's strengths against itself. And these are the different tactics inside of Brutal.
But I want to leave you with one bonus feature. And this bonus feature is something where say you want to always get critiqued all the time. You want to make sure that the AI is always giving you critical feedback irrelevant of what conversation you're having. Well, you can set up system preferences inside of CatchBT Claude and Gemini where it always looks at a very specific prompt. So, if you're inside of chatbt, you can go to your settings. You go to personalization, then you go to custom instructions. Then, inside of there, if you want, you can copy and paste these instructions in there, and every conversation you have with ChatBT, no matter where you are, it'll give you critical feedback. The same thing applies to Claude and Gemini. They have very similar settings. And here's the prompt. Prioritize substance over compliments. Never soften criticism. If any idea has holes, say so directly. Then we say, then we give an example of how that should look here. And then right after that we say challenge assumptions, point out errors. Useful feedback matters more than comfortable feedback. So we're basically telling the AI this is the type of persona we want you to have going forward in any conversation. And note this is pretty intense, so I don't recommend it for everybody, but I wanted to give this guidance out there just in case you were interested.
That being said, let's recap everything we've talked through. So for Brutal, we have six steps. Starting with begin fresh. This is where we're trying to wipe the AI's memory, starting in a fresh temporary chat or an incognito mode to prevent it from having understandings of what our preferences are. After we've done that, we need to choose the right model so that each model has its own type of training and some models are more honest than others. So, if you want super honest feedback, you can just go to Grock from XAI. If you're already baked into using Gemini or Chatbt, that's when we move on to you, which is using a very specific persona. So, the three levels I provided you of intensity are devil's advocate, red teaming, and Ramsay. You can choose whatever persona you want, but here's just some ideas and prompts you can start with. After you, we move on to T, which is disassociating yourself from the idea, because remember, the AI is trying to protect your ego, but not anybody else's. So, if you detach yourself from the idea, you're likely going to get better feedback. After that, we have asking specific questions. So, if you ask a vague question, you're going to get vague feedback. So, we want to be specific both on the question and also the type of critique that we're seeking. And after all of this, if we're still struggling to get honest feedback from the AI, we can actually turn the AI on itself doing a bit of a jiu-jitsu move and have it grade its own feedback and give it a specific ask to critique the feedback to get something better. And that's it.
So, if you enjoyed this, a quick reminder of two things. First off, Blow is a 30-day AI insight series, completely free. You'll get 30 insights in your inbox so I can apply AI to your business and your work. The second thing is if you'd like to work with me, Blow a series of offerings to see if there's a good fit between the two of us.
Okay, so now you know you can get AI to tell you the truth, but which AI features should you learn in the first place? I ranked every single one of them. Some will save you hours every week, but also some are a complete waste of your time. I'll show you which is which right here. So go ahead and click that video. Go ahead, click it. Click it. I'll see you next time, internet.