That big detailed prompt you've been refining for months, that has a persona for the AI, maybe 15 instructions or a step-by-step recipe, that's quietly making GPT 5.6's output worse, not neutral, worse. And here's why. OpenAI just published their official prompting guide and it contradicts most of the advice from the last year. The newest models are trained to figure out the best path on their own. When you spell out every single step, you're forcing a smart model to follow an average process instead of the best one. If you're new here, I'm Dylan. I run an AI consultancy. And I see these oversized prompts costing clients hours every single week. The fix is five parts. I'll walk you through all five and build a real prompt with you piece by piece so you can rebuild your most used prompt in about 10 minutes. And this is Rocky. Let's get into it, internet. So here's the blog post that OpenAI just published only a few days ago. And in this blog post, it covers the different prompting tactics that are relevant for their newest model, GPT 5.6. All different sizes, small, medium, and large. And most people have no idea this blog post was even published and very rarely do they read it. And my goal is to synthesize this blog post into five critical tactics that you can apply to prompting this model going forward. And many of these tactics are actually applicable to newer generation models such as Fable 5 as well. And to make sure that these tactics are concrete, I'm going to walk you through a single example along the way, but note this is just an example of me being able to convey something to you. You can apply these tactics to any type of prompting or any type of use case. So our example here is taking a meeting transcript from a client and converting that into a status update for that client. And doing it with the best practices for GPT 5.6 specifically. Now, the very first tactic is something I've talked about quite a lot in the past and it applies to all new generation models, which is instead of focusing on the steps, the processes, and the recipe of your prompt, you need to focus on the destination. Because many prompts in the past, the older prompts for the older models, they tend to include a lot of steps inside of the instructions Saying, do this, then do that, then do this, then do that. This is constraining on these really intelligent models. We're getting in their way because oftentimes, there's way more steps than just the four we've listed there. There's likely 40 if not 100 steps AI has to take, micro and macro, to achieve the task at hand. And it often knows a more effective path to that destination. We may give it path A, but path B may get us there much faster, but also at higher quality. So, what we want to focus on is what we want in the end. So, focus on the destination, not the process. And the way that we can do this is by simply specifying the outcome and the audience if the audience is relevant. So, let me show you a simple example here. So, the first prompt here is a bad prompt. It's talking about the old way of prompting. We've listed out all the steps here. We said, I want you to summarize the transcript. First, read through it, then pull out the key points, then organize it for me, and then write a summary. Quick pause. If you're enjoying this, you're going to enjoy two other things. First off, below is a 30-day AI insights 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, below are a series of offerings to see if there's a good fit between the two of us. Now, let's get back to the video. All of this is unnecessary now. Instead, what we can do is we can specify simply the outcome and the audience. So, here in our case, where we're taking a meeting transcript and putting it to a client status update, here we're saying, I want you to turn this meeting transcript into a short status update for my client. Put the decisions and next steps first. So, the two things we called out here are the outcome, so this is a status update, and then the client is the audience. And we even then added at the back half some preferences on what we want the structure look like for that outcome. So, we weren't specific on the steps, we were specific on how the outcome should look and what good looks like. And that point right there on what good looks like is a critical thing to add as well if you can. So, if you have an output that looks good, you can share that with the AI, it'll quickly get to that destination for you. Now, a quick caveat I'll add to this tactic is that sometimes the steps do matter. It's very rare that it does, but sometimes it does. So, if you're doing something in relation to audits or compliance, then you may need to actually specify to the AI certain steps it needs to take in which order because the output itself is the deliverable. Because you're going to use this log
