[00:02] they actually remove over 80% of Claude Code's system prompt for more advanced models. Now, these are based on the new Claude 5 generation models. It's also super useful to learn this stuff if you are using any sort of AI agent. So,
[00:16] take to this to really understand how to get the most out of your AI agents when these are the new rules of context engineering for Claude 5 generation Crazy thing about this is they've removed over 80% of Claude Code system.
[00:32] And they actually found when they removed that fluff, they got more out of these models. And by doing that, they actually didn't see any measurable loss the 80/20, but they actually got the same results, which is crazy in itself.
[00:46] So, this is all about something called context engineering. It actually makes a big difference when you're using Claude Code or any sort of AI agent. And the thing to note here is like much of the context assembled from the system
[00:58] prompts, the skills that you have, Claude and D files, the memory, other sources, this bloats up the context you actually have. So, for example, when you send a message to Claude, that message is only a small piece of what Claude
[01:11] Claude also gets a system prompt, your Claude and D files, skills, memory. And context. Now, you want to think of it like a lunchbox for someone. Say your message is a sandwich, but the lunchbox also has napkins, has a note, has a
[01:25] fork, has snacks. And context engineering is deciding what goes in the whole lunchbox, not just a sandwich. So, the tricky part is your message changes every time you use Claude Code, but the context is reused for loads of different
[01:38] messages. So, it has to be general instead of specific. Now, what they actually found is that unhuddling Claude is you just remove a lot of the constraints that it has. So, basically
[01:52] less is more here. So, if you have loads of rules, like for example, leave documentation as appropriate, or for example, do not add comments to the kind of clashes with each other, especially if the user of Claude code is
[02:06] example, you're using AI agent, and let's say for example, that agent has loads of rules, but then when you prompt it, sometimes your prompts actually contradict the rules that it has. The problem with that is like Claude or any
[02:20] you're trying to do, but it has to think more carefully about overlapping, and also these conflicting messages it has. And so, these used to be really useful for reducing like the worst-case scenarios, but actually these days you
[02:34] with the newer models. And also, there's tons of more tools inside your AI agent. So, for example, if you look at Claude code, you used to rely on Claude.nd, as guidance, but now you actually have an inbuilt memory,
[02:48] and artifacts, and skills. And so, Claude has new ways of loading the context without using the Claude.nd file. So, instead of your instructions fighting each other, you can actually just use the rule less is more,
[03:01] just use the rule less is more, and over 80% of Claude's system prompt can be removed, especially for the new models. Older models need stricter and stricter rules. Newer models have better judgment, so many rules can actually be
[03:13] five different rules you can see here. So, give Claude rules was the old way, Before you'd have to give Claude examples, now you can design interfaces. You can put it all up front, or you can use progressive disclosure. And this is
[03:28] the old way, this is the new way. So, before you'd have to repeat yourself, So, before you were giving Claude loads of rules, but now you can actually not really worry about that too much, because instead Claude can use judgment.
[03:42] prompt of Claude code, it's like, okay, there were hard rules like write no The problem is sometimes it's just wrong, like some users want documentation, some complicated code genuinely needs a larger explanation.
[03:55] And so the new instruction is basically write code that looks like the code around it, matches new comment style, naming and habits. So instead of a rigid rule, Claude can just understand what code you've already got and then use
[04:08] you'd have to give Claude examples, but they actually found with the newest models like Claude Opus 5 and Fable 5 is that giving examples actually constrains
[04:20] them to a certain exploration space. So before this rule for Claude of like show Claude examples how to use them, they kind of create fences where Claude would stay inside the box. Whereas with these new tools, scripts and files, basically
[04:36] Claude has parameters that allows it to be more expressive. So before it'd stay in the box like this, whereas now it has these different status options. Now up front. Now it uses something called progressive disclosure. Basically
[04:50] times. Now you might be wondering okay, how can you apply this to your own usage of Claude for example? Well, you can apply this to for example files. So people before they would have like a central of like every node entity
[05:05] otherwise. But now what you can actually have is a tree of files that can be loaded at the right time. Because like most of the time the information you have before was just kind of sitting there taking up space, right? Now you
[05:17] can have like separate skills for each workflow and Claude would just open them you don't need like everything inside Claude MD as one giant book containing skill files and then each time for example it does SEO, it uses the SEO
[05:32] skill. Each time it uses front end, it's going to have a skill for front end to repeat yourself, especially like repeated instructions. This is really conversation with Claude, sometimes it would get lost on the context and not
[05:46] find his way back. And he would get completely off task, if that makes sense. But now, you can actually just put instructions on how to use tools in actual system. So, what this means basically is like older models sometimes
[05:58] of a long context. So, people had to keep like copying and pasting the system prompt, you'd have to do it again and again and again just so it doesn't go off task. But now, you don't need to do
[06:10] for a tool in the tool description and delete the copy from the system prompt. Something else that people used to use was memory in Claude MD files. And so, what they actually used to recommend was saving things to Claude's memory by
[06:23] using the hashtag hot key to write to their Claude MD file automatically. But automatic memories. And you can also plug in Obsidian. So, either way, you don't need to say like save this. Or for example, add this to your Claude MD
[06:38] file. It will remember the simple pieces as you go along with auto memory. So, something we're still using in Obsidian. Uh, is simple specs. You would have like a markdown file and those markdown files would help Claude have a reference point
[06:53] Particularly if it needs to store something in like code base for later. But now, actually, what you can actually use because they has the new they have the new artifacts feature, you can reference HTML artifacts. So, you can
[07:07] code. You don't need to use a markdown file. So, that could be like a HTML artifact you use instead of a plain markdown file. You can eat code as a spec. So, a detailed test suite is a spec and so is a function in another
[07:20] code base. And also rubrics. So, they basically teach Claude your taste, which is really important actually because like Claude can basically code anything. designs that he creates, often they're quite bad. And that's because he doesn't
[07:33] have taste. So, rubrics allow you and allow Claude to try and verify the taste the verify agents with those. So, if you what you got the system prompt and this is only for people building their own
[07:47] prompt, that's only something you need to edit if you're creating your own harness, but the Claude MD file is pretty important. So, if you're using Claude code particularly, you want to keep it lightweight and just briefly
[07:59] explain the stuff that might surprise Claude. Like for example, all our types Right? So, you won't write anything obvious inside the Claude MD and you
[08:11] keep it simple lightweight. Instead, what you want to use is progressive like make a verification skill and just point it and point to it from your you're not using skills in the city it's for. So, if you run a workflow like for
[08:25] example, design a page, you can say, "Okay, save that as a skill in the guides to let Claude find information when it needs it. Now, again, you don't want to have too many rules inside there. That's actually something that
[08:39] I've probably made the mistake of a nature previously in the future. We've our skill files. So, if you are having bought new skills inside the workflows lightweight. That's actually something I plan to use and just reduce the blow.
[08:54] also for references, this again leans into progressive disclosure, which is instead of having everything inside the Claude MD, you add context later. So, you could use app mention files to include in those references and then
[09:07] Claude can use those later. So, you can go into Claude and then tag in the file. mock up, whatever you want. And again, it doesn't need to be like a markdown file. It could be a HTML mock up. It could be for example a a HTML. And
[09:21] you can use for references instead because it's super precise and Claude is really fluent in code. So, for example a HTML mock up of a design will usually be a written description or a screenshot of that design. And then finally, simplify.
[09:35] the bloat inside your skills. That's something I plan to do. System prompts, of them a lot more. And that just reduces the constraints and gets more out of the Claude 5 models you're using. So, thanks so much for watching. That is
[09:49] the new guide from Claude and what they've changed inside Claude code recently to just help you get the most out of Opus 5 and also Claude 5 and 5 and any of the new models that come into the 5 category. If you want to get more
[10:01] etc., you can go inside the AI Profit Room. Link in the comments description or go to the AI Profit Room.com. This is my community that will save time, grow, can post questions inside the community. It's a great group of people where you
[10:14] I personally answer the questions inside there, and so does everyone else. And help you out if you need to. Inside the classroom, you get access to all of our beginner, you can go from beginner to expert with the 6-week course. If you
[10:26] advanced stuff, you can get that right here. If you want our Agent OS system that allows you to orchestrate all of your AI agents called Hermes and stuff like that, you can grab that. And then also, we have tons of trainings, as you
[10:39] weekly coaching calls to get help and support real time, share your screen, you can meet other people using Claude and other AI agents locally near you. message me anytime you need. And we actually have over 204 pages of
[10:56] learning and growing with this stuff. So, it's just an awesome community, great people, lots of people growing and learning and scaling with AI automation. comments description or just go to the AI Profit Room.com. Thanks for watching.