Transcription
What if you could get better results from AI with a simple prompt like this than with a complex prompt like that? Well, the reality is you can because getting the most out of AI is no longer about better prompting. The secret lies in giving it better context. But what does that actually mean? And how can you use it to transform your AI from a forgetful assistant to your most valuable team member?
Let's start with the foundation of it all. Every AI that you work with, every LLM has read billions, even trillions of different documents. It has the equivalent of a business degree. So, it does know some things about business. The problem is that it knows nothing about you. Let me demonstrate what I mean.
I use an app called Whisper Flow, which allows me to speak to the LLM instead of typing, which is much slower. So, I'll go ahead and hold down the function key to activate it. "Give me a marketing strategy for our flagship product."
The AI just charges ahead and gives me this what looks like a very well-thought-out and comprehensive strategy, but it's really not. It doesn't know anything about my audience segments. I mean, it's guessing here. This is not really correct. Okay? It's not at all about consultants and agencies. It's not about operations managers. It's about business owners. We actually don't do anything with SEO. So, that's wrong. We're not really that big on partnerships. This is all just basically made up and and invented pretending like it has some idea of what we're trying to accomplish.
The reason this response is not very good is that it's based on only level one of context, which is the training data. I always see people obsessing about using the latest, most advanced model because it has this many trillions of parameters instead of only that many trillions, which is just insanity. That makes no sense to me. What matters now at this moment is not the size or the sophistication of the model. What matters is how much context you give it. That's why we're going to be focusing on that context for the next four layers.
Did you know there's a 120-page document controlling and shaping every response that an LLM gives you? Most people out there have no idea this even exists. Level two here is called the system prompt. Every single LLM out there has one. These system prompts are actually created by the companies and the teams that make the LLM. The important thing to realize with a system prompt is that you can't edit it. You can't change it, but you can exploit it.
Here's what I mean. Claude actually has seven different trigger words that if you use them, it triggers it forces the LLM to do deeper analysis and deeper thinking. By using any of these seven words, or especially more than one, you can easily access a 10x or more improvement without any further context. So, watch this. I'm just going to go and copy the same prompt as before in the same chat. I'll paste it again, but this time I'll add just a few words such as "give me an in-depth, comprehensive marketing strategy for our flagship product." So, just added really two or three words.
Let's just go through this briefly. I want you to notice that this is actually a deeper, more comprehensive market strategy. So, it's taking a step back and thinking about things like the market size, whether the AI training market is growing, which it is. What percentage of business owners are interested in such training? It's looking at the competitive landscape. Notice that it did not do this before. It started with core positioning. So, it's going deeper. It's going wider. This is not just a longer response, and you can see it's still working on it down here. It is deeper. It's more thoughtful. It's actually thought about the implications more. It's It's actually a better result.
So, I want you to notice that I still haven't really given it any extra context. All I did is add two or three extra words to the same prompt, but what I get back is a response that is longer, but also better, deeper, more thought out, and more strategic.
By the way, I put together a breakdown of both Claude 4 and GPT-5's hidden system prompts, including trigger words, the secret behaviors, and even the design principles that they follow when creating responses. Just scan the QR code on screen or check the link in the description.
So, we've already made a lot of progress, but these hidden features are really only just level two. Watch what happens at level three, which is the first level where you can finally take control. I'll go ahead and paste in the same prompt as before and just add "an online training program that helps business owners implement AI across the major functions of their business." Now, notice I'm barely adding any more to my prompt. This secret sauce is something I've changed in my settings, which I'm going to show you in a bit.
What we get back is a similar response, but with a few key differences. One difference is that it's using almost exclusively bullet points, which is my preferred format in what I get back from an LLM. Second is it's telling me its level of certainty. This is something I love is I actually have it tell me, "Is this something that you're highly confident in? Is it something that you're not very confident in? What would change your mind or what are the weak spots or what's missing from your analysis?" Which it can do. So you can see for each main section it's said whether it's high certainty, moderate certainty, low certainty, the highest certainty. And finally, this is kind of hard to detect, but it's become much more conservative in its estimations. Instead of giving me these pie-in-the-sky kind of unbelievable claims and statements and predictions, it's being much more conservative in how it's thinking.
What you're seeing happen here is level three, which is the user preferences. You can tell the LLM how you want it to communicate with you. You can set those communication preferences once and they'll apply everywhere all the time. To find this, you'll want to go into the preferences or the settings of your LLM. Here's what it looks like in Claude. And then in Claude, it's called personal preferences, which you can see in this box right here. And I've only given it three, which are the three that we saw: "Answer in bullet points and sub-bullet points. Be conservative in your thinking. And tell me your level of certainty." Once I've added these preferences into this box and hit save, any new chat that I create, I don't have to ask for it to do those things time and time again. It already knows those personal preferences apply across every single chat that I have with Claude. And most other LLMs have something quite similar. For example, in ChatGPT, the same functionality is called custom instructions.
But here's where it gets even more powerful. Level four is project knowledge. And this is really the most transformative one, the one that is so completely underappreciated by almost everyone that I work with and that I see. Let me start by demonstrating it. I'll go ahead and paste in the same prompt as before and then hit return.
Right away, I notice a few things. First of all, it knows the name and the abbreviation of our flagship product. So, notice up here, I did not tell it what that product was. It already knows. It also knows that we're in cohort two, which is a detail contained in the project knowledge. And then as I look through the marketing strategy that it's given me, I notice that it's just, for example, it knows the exact range of business sizes that we are serving, which is 10 to 50 employees. It knows that we work with a lot of digital agencies and professional services firms, including in non-tech industries. It has the names of previous participants in our program and their testimonials, the nice things they said about their experience. It knows that we're thinking of regularly publishing content, for example, via an email newsletter and LinkedIn, which are the exact two platforms that we're most interested in to kind of keep people interested and engaged in our ideas. It knows that we use Luma for our events, uh, for people to sign up for. It knows that we have a direct sales team and that we use email marketing as our primary channel. It knows our competitors and who we are positioning ourselves in comparison to. I mean, what you see here is something that is so far advanced beyond what I what I think most people are using LLMs for. It is taking advantage of a tremendous amount of detail and specificity which is contained in the project knowledge. The AI that I'm working with now knows my products, my pricing, the metrics we use, everything related to our flagship product.
Let me show you how this works. I'm going to click over here where it says projects in the sidebar. And you can see these are like different workspaces, each one dedicated to a specific project. So this is the one that we were just in and it has a series of documents that are all related not just to this program, which is Second Brain Enterprise, and not just related to cohort two, which is the next one coming up. But even within that, it's really specific to sales and marketing. So you have things here like the applications of previous participants, what they said they wanted, what they said their pain points were, excerpts from customer testimonials. So people who have already been through the program, what did they get out of it? And then this third one is actually a download of a big Google Doc that we've compiled that has all the marketing strategic positioning and the content pillars and the messaging and the differentiators and everything you could possibly want when it comes to marketing. So the way it works is that anything you upload here to this area called files or project files is drawn on for every single conversation that you have within this project. So if I leave this project, that's different, but any chat that I initiate here automatically makes use of and is aware of everything here in the project files. Basically, every conversation that I have with AI within this project is like collaborating with a fully onboarded, highly experienced team member versus an intern on their first day.
Now, watch this final level, which brings everything together. You've built four layers of permanent context. But now comes the moment of truth. "Help me create a new prompt that will generate a comprehensive, in-depth marketing strategy for cohort three of Second Brain Enterprise, which will take place in October 2025. Do an analysis of all the complaints as well as the wins and the benefits that past participants have reported. Include assignments and OKRs for every single member of the team. and propose at least 10 ideas of YouTube videos that we could create with a direct connection to our persona and the benefits that we offer them."
So, notice what I'm doing here. My goal is to have a prompt that I'm going to use to ask Claude to generate a more specific, strategic, and actionable version of my marketing strategy. But you actually don't need to save this library of comprehensive prompts that you found online somewhere. You can get the AI's help in creating the prompt in the first place. So that's what I've done here. And I'll go ahead and hit return.
All right. So you see it has drawn on the project knowledge, the project files to you already really start to answer and outline what some of those may be. And then it's giving me a comprehensive prompt in this kind of special window that will in turn be used to generate the full marketing strategy. Okay, so you can see it's quite extensive. I'll go ahead and click the button right here to copy and then head into a new project. Paste that prompt that it had just produced for us and hit return.
Now this is really the apex of leveraging an AI for intelligence. Notice that it has a lot of details about the nature of our program, including key criteria that are part of what we offer, including growing your company without adding headcount or sacrificing quality. It knows what our differentiators are, that we're industry agnostic, that we do implementation in real time, that we're focused on organizations. It has access to all of our proof points via testimonials, and then it has all the context to provide a series of YouTube video ideas that are actually quite good. Knows about the master prompt. It knows that we're focused on small teams trying to make a big impact. It knows that we're positioning ourselves as a much faster, more effective alternative to DIY AI training. And then it's even gone above and beyond and given us case study templates, email nurture sequences. This is part of that comprehensiveness that we asked for. Has sales enablement material that our sales team can use, discovery questions, a framework for calculating the return on investment of the purchase, how to handle objections, team assignments and OKRs, as we requested. This is really honestly, it would take me weeks. It would probably take me two or three weeks to get to just this level of quality and thoughtfulness that we got to in a matter of minutes.
This is level five, your actual prompt that you put into the chat box. What I want you to see here is that just because we have context engineering doesn't mean that prompt engineering, which came before, is obsolete. What this is demonstrating is that when you have the right context preloaded and already in place, your prompting is vastly more powerful and even more effective.
Now, I want to leave you with an exact road map to what I call context mastery. In the next 10 or 15 minutes, Google "your AI plus system prompt." There you'll find those trigger words and other hidden features for the LLM that you use. Tomorrow morning, add three or more user preferences. Just think about how you like other people to communicate with you and ask your AI to do the same. By the end of the week, create one project with your current work files. Those could be text documents, Google Docs, cloud storage drives, wherever your documentation and your files related to a specific project live. You can attach those or connect them via an integration or, worst-case scenario, download them and upload them into the project files. That's it. Three actions within one week at most. Then watch your one-sentence prompts outperform what previously required an entire essay worth of prompts. Watch this video next where I show you the master prompt method, the exact template we use at Forte Labs that turns these context layers into what I believe is our most important competitive advantage as a