For about 2 years now, we've all learned the same skill with AI, which is to push it. Longer prompts, more steps, anything to make it do enough. But the new Claude, Fable 5, has the opposite problem. It does too much. It'll run for hours, take actions that you never asked for, or even claim it finished work it never touched. If you're new here, I'm Dylan. I run an AI consultancy. And after years of coaching people to get more out of AI, I'm now teaching them to rein this one in. I'll give you the three controls I'm using with my clients, plus one sentence that makes it prove the work that it did is real. So, let's get into it. Now, when choosing to use this next generation of models, such as Fable 5 or GPT 5.6, this is the problem that a lot of people are starting to face. The old prompts that worked with the old generation of models, like Opus 4.8 and GPT 5.5, to get the most out of those, we would have long prompts that had many steps, many constraints, and preferences that would get good quality outputs. But now that these other models, these new generation of models that are so intelligent, we actually need to rein them in or control them so they don't go overboard on whatever task we've given them. And also, we don't want to get in their way. Now, before I talk about the controls and some of the things that we're going to change when we interact with these new generation of models, note that Fable 5, after a certain date, I think it's July 7th, is going to be paid to use. So, it's not going to be baked into your subscription, irrelevant of whatever subscription you have. So, if you're on Max, Enterprise, doesn't matter. You have to pay to use it. Which means I wouldn't recommend using this for all your tasks, only the most important tasks that are very complex and you want to have a high-quality outputs associated to it. And I'm sure this will change over time. Um as Anthropic gets more compute, they'll probably bring this model back into the subscription offering, but for right now, it'll be paid to use. So, just be aware of that. Now, with that caveat out of the way, when we prompt this model, what are the things we need to consider? Well, the first thing we need to do is take our old prompts and reduce them drastically. Reason being, and this is actually directly from Anthropic, they have a user guide for Fable 5, and in that user guide, they explicitly state that old prompts that worked with older models often are too prescriptive and can degrade the quality of the output for the newer models. And that's because they are so intelligent and all the scaffolding and prompting that we did for the old ones isn't needed. And they're not just preaching this, but they're actually practicing it. So, for their tool, Claude Code, which is basically a technical version of Claude Co-work, they cut 80% of the code in the prompt behind the scenes for these newer models. And that's because all of that cruft around the AI was holding it back. And what they call it at Anthropic is un-hobbling the AI. We're getting out of the way of the AI because it is so intelligent that we want to avoid constraining its true power. So, this is just showing you how they're doing it, Anthropic, with this new model. Now, probably the most important slide of this entire video, if you paid attention to nothing else, pay attention to this. When you're prompting these new generation of models, Fable 5 and GPT-5.6, you want to give it three things. Quick pause and regular programming. This video is brought to you by me, as always. Two quick 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, below are a series of offerings to see if there's a good fit between the two of us. Now, let's get back into the video. You want to give it a goal, the audience you're serving for that goal, and most importantly, what done looks like. Now, this template here is from Anthropic in their user guide. You can copy and paste and use this for yourself. And the important thing about this prompt, and these three elements that I just walked you through, is what we're doing with the AI is we're showing the direction of where we're headed, but we're not telling it every single step along the way. Because the AI is so intelligent that it'll likely have a better idea of the steps needed to get to that destination better than we will. So, all we're doing is pointing the AI in a direction, telling it what done looks like, and then saying and And that's what this template represents. So, in this prompt, we're simply telling the AI, "I'm working on blank." So, this is the goal, the big task. "And I'm working on it for fill in the blank," which is the audience. After that, we add some additional caveats, such as they, the audience, need additional things, such as fill in the blank. So, if there's additional