Anthropic, the company that brought you Claude, just deleted 80% of the instructions that they'd written for their own AI. Not because any of it's wrong, it worked for years, it's just getting in the way now. And the stuff that they cut is exactly what I've been telling people to add. More examples, tighter rules, saying the important things twice. If you're new here, I'm Dylan. I run an AI consultancy. And I've spent this week undoing my own advice inside of client setups. And here's what makes this hard to catch. An out-of-date setup doesn't look out of date. Your projects and skills, they still run. The output still looks fine. But you wrote all of it for last year's models. And the newest ones, they don't need that help anymore. So, let's walk through the three habits that changed, and I'll give you a prompt for each one that goes and fixes what you've already built. So, let's get into it. And before we get started, here's the blog that I was referring to. So, Anthropic just put this out probably a day ago, talking about new generation models. So, models like stable 5, Opus 5, GPT-5.6, etc. And here are the three pieces of advice that they called out as myths. So, in the past, we were often told that it's best to share examples with AI, so it knows what good looks like. Or to give it hard rules, so it follows them in a specific way. And even to repeat ourselves in different locations to make sure it follows the really important things. Maybe repeating the same instruction at the top of a prompt, at the bottom of a prompt. And even adding it to the chat as well. All three of these things have been labeled as myths from Anthropic. And it's important to know that it's not just the Claude thing, but it's also a ChatGPT thing. OpenAI released GPT-5.6 only a few weeks ago. And when they released that, they showed a series of studies where when they tested the AI with a shorter prompt on the same exact task, they found that it wasn't only just more effective, so it scored better on different benchmarks, but also was more cost efficient. So, it saved roughly 50% on cost for each of those tasks. And it's specifically for those new generation models. Now, let's start with the first habit, which is sharing examples with the AI. Now, it's not that we shouldn't share examples, but when we do share examples, we need to focus AI's attention on a very specific piece of that example. What do I mean? So, when you give an example to an AI, you simply ask it to match this example for whatever task going forward. When we do that, we're actually setting a ceiling on the AI's ability to unlock its full intelligence for that specific task. Because when it looks at your example, it narrows its focus on mapping how you've done the thing in the past. So, if it's a report, it's going to use the same angles and approaches and ways of thinking that you used in that report going forward. But, there may be many better ways of doing that task, but we've limited the AI's intelligence on that specific way of doing things. So, we need to be careful when we share examples with them, so we're not limiting its intelligence. And the way that we can do that is by understanding that an example has two specific jobs. Quick pause. If you're enjoying this, you're 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 if I 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 go back to the video. One is when we give an AI an example, that first job is the aesthetics. So, how does it look? The second job is the actual answer. So, what is our approach to this specific example? The aesthetics are fine and are still the reason as to why it's good to share examples with AI, because we want it to match to our aesthetics. So, our colors, our fonts, logos, etc. But, when sharing examples to an AI, we don't necessarily want to bias its answer to how we've answered in the past. So, our approaches, our angles, our options, etc. All that stuff was something we did in the past, they might have better alternatives for that specific approach or that specific task at hand. So, this distinction's really important for this specific habit. So, I'll give you two prompts in regards to actually creating net new prompts and skills with this habit in mind. And then I'll give you another one for fixing your old projects and skills with the old way of doing things. So, start with new first. And this is tackling the standards things. Remember we had our approach and our aesthetics. We're going to start with the approach and then we'll talk about the aesthetics in a second. So for the approach, what we want to do is we want to share an example with the AI, but this is an AI that's not going to have the skiller project baked into it going forward. This is just ex

This is what a great onboarding process looks like with AI. The second that a client says yes, the contract is already in their inbox. It is e-signed and countersigned before they can even second-guess it. Now, just a minute later, a welcome packet lands. So, what happens next? Who does what and when it is done? So, they feel certain that they picked the right team instead of wondering if they wasted their money. And from there, their first invoice goes out and the moment that it is paid, their access and permissions turn on by themselves with nobody adding them by hand or doing it manually. And now, the kickoff call schedules itself from their calendar and before you join, AI has read their intake answers and handed you a one-page brief. So, now you can just walk in already knowing their business. From there, everything lives in one client portal. So, every document, every status, every invoice, so you never get another status update email from your client. And then, every month, a report writes itself and lands in their inbox. What got done? What is next without anybody having to pull it manually? And then, a check-in quietly asks how it's going and if the answer is lukewarm, it flags you before they ever think about leaving. I've built this exact flow in over 60 businesses like agencies, clinics, home services. It does not matter. This is the easiest way to keep your clients from churning and increasing their lifetime value. Comment "Need it" and I'll send you the full setup guide.