Transcription
There are hundreds of clawed code tutorials out there, but all of them are ignoring the most important factor. The reality is if you, the person watching this, doesn't use the tool properly, then nothing else matters. So, today I'm bringing down these six fixes you need to make to 10x your output. Each fix comes from problems that I've identified after working with hundreds of business owners and their employees.
So, fix number one is embrace your attention limitations. In the past 3 years, the ability for AI tools to create productive output has gone nuclear. Yet, people's attention span has only gotten worse. So, this divergence is the problem. Simon Wilson put it perfectly on Lenny's podcast >> sort of personal skill that we have to learn which is finding our new limits like what is what is a responsible way for us to you to to not burn out. >>
So how do you embrace your own limitation and set yourself up for success? First you need to understand where AI excels and where you excel. AI is great at fast continuous execution pattern matching at scale whereas you're great at judgment taste and knowing when something's actually finished. With this in mind, put your task into high and low effort where you want to put the intellectually complex stuff in your freshest hours. Then put the simpler, repetitive stuff at the end of the day. My system for this is dead simple. Every day I have an AM slot for complex tasks and PM slot for chill stuff that I can do while watching Netflix.
The second is that you want to consolidate your notifications. Different tools ping you in every direction will quickly draw down your attention. So I set up my workspace so that Slack is the central hub for my entire business. Every clawed automation and loop that I run reports back through Slack, reducing my context switch. I found that the easiest way to do this is using Slack web hooks and having specific channels for specific automations.
The third is that you want to have Claude meet you with where you're at. Use Claude's Dispatch to access your computer's file system from your phone. There's no need for you to sit at your desk all day. You can just do these things remotely. Also, a key factor here is use voice first tools like Whisper Flow or Hex so you don't have to type text in. voice keeps it so that you're not typing all day, which will just burn you out.
So, attention reclaim is fix number one, but the next fix is an issue that 99% of people run into. Fix number two is quality over quantity. Since AI started taking off, producing things has become essentially free. So, just getting things done in a world of abundance is worthless. But what is still valuable, quality? Let's think about Nike for a second. They've spent decades building their brand. And every piece of content that they publish either improves the brand or it hurts it. There's really no neutral. So, if they just had 3 months of AI slop, then the brand would be cooked. And the same can be said for you. Yes, you can produce millions of things and hope that one sticks, but your personal brand equity will suffer in the process. And I'm not just talking about content creation. This matters even if you've never created a piece of content in your life. People notice when you send them AI emails. They notice the AI generated reports that aren't thought through. That's that quantity that you're just producing things and it doesn't provide value.
So, how can we leverage AI to increase quality instead of focusing on increasing quantity? We want to AI time travel. And we do this by building a system that gives you feedback before the finalized version is ever submitted. This makes it so that you can take that feedback, improve it, and then the final version is that much better. For any highle strategic task, you want to create a skill called ask the board. Find four or five thought leaders in your vertical with publicly available content. That's going to be your board of directors. For example, let's say Alexi, Mark Cuban, Andrea Karpathy. You would use this prompt to then scrape their public content and create their clones. Now, if that sounds like a lot to you, you can use my plug-in buildpartner.ai and run this, which is already trained on the exact people I mentioned.
Now, for anything that's customerf facing, you want to create an internal focus group skill. Have Claude interview to create a clone of the end user of whatever you're working on. Here's a prompt to help extract this information. And this applies really to anyone, right? A person marketing a $10 product on YouTube to a person presenting a financial report to investors. Trust me, this stuff really makes a difference. I used something similar at my last startup which raised over $20 million. And the third way is for any day-to-day work task, create a skill called clone your manager. Aml, who runs growth at anthropic, cloned his manager from her blog post, her slack messages, and her emails. Now he can get feedback from her before actually submitting anything to her. So as AI makes quantity of production go parabolic, quality is at more of a premium, not less.
Now, before we get to the next fix, which is universally agreed upon, you may find that a lot of what we're covering feels a bit technical, which brings us to today's video sponsor, Nexus AI. Nexus AI is a AI platform, which is a category of tools that I love. The reason being is that it removes all the technical skills that are normally needed to set up an AI agent system, and it can save you hundreds of dollars a month because you don't have to pay for five different AI tools.
Now, there are three specific features that I love. The first is that every flagship model is under a single Nexus login. So, ChachiBT, Claw, Gemini Grock, and 200 others can all be accessed at once. And this is important because there's really no single best AI model. Different ones are better at different things, and Nexus will automatically route the specific prompt you're using to the model that's best suited. The second is that it's all combined under one price. Instead of having to go out and pay for each tool separately, it stacks them all together and can save you up to $200 a month. And if you're watching this, you'll actually get 50% off Nexus, which is pretty dope. The third is their no code AI agent builder. If you can use the internet, you can use Nexus to build an AI agent. You just describe what you want and Nexus will build the agent for you. And whether that's for one-time jobs like I need to do deep research on this specific topic or repetitive tasks like turning your notes into a weekly report automatically, Nexus will handle this for you. For me, I have a lot of products in my ecosystem for my YouTube, my newsletter, build partner, so it's hard to keep track of. I can create a weekly report that it analyzes my entire marketing funnel, connects to my Slack, and then sends an update message so my whole team can be on the same page. This is a task that I normally would never do just because I don't have time to do it.
So, to check out Nexus, click the first link in the description where you'll get a discount of 75% off annual plans. And to be honest, I'm not entirely sure how long this discount will last because I had to really push for it. So, go check that out.
Fix number three is stop prompting, start planning, and empowering verification. Generally, I bucket people into two groups, sculptors and gardeners. The sculptures hands-on, crafting the final product step by step. They prompt, they wait for the response, they adjust it, they prompt again, they grind until that final product is ready. The gardener, on the other hand, position plants to grow, it waits, then it harvests the fruit at the end. So, they create a plan, they establish goals, and they make it clear how AI can verify the output. To build 10 times faster, you need to move from a sculptor to a gardener. I have videos on my channel where I dive deep into this, but the short version to become a gardener is we want to first plan, then have AI execute for us, and then use AI to verify the output. This general framework will allow AI to complete tasks efficiently.
So to plan anything, the simplest way to do this is /pl I want to build task. Interview me to identify any gaps in my knowledge. And if you want a more robust skill, once the plan's created, you have AI execute it and then verify. If you get the plan right, execution is pretty straightforward. The differentiator is verification. Simply put, if you can have AI verify that the final product is correct, it'll know when it's incorrect and it'll keep going until it's correct. Enthropy actually said this directly. Verification has had the most measurable impact on Claude's output quality internally. It's their highest single quality lever in their entire system.
There are three ways to integrate verification into your work. The first is update your Claude MD with explicit verification language. Just add one sentence before returning any work. Verify that it works and the task is complete. If you can't verify it, fix it and rerun. That single line will force Claw to check its own output before handing it back to you. The second is you need to wire up MCP servers for the external systems you actually work with. So for whatever systems you're using, they likely have an MCP or an API connection that Claude can connect directly to. By establishing this connection, you make it so that Claude can get realworld data, not just the data it's seeing on its computer. If you type /mcp into Claude code, you'll see what's already connected. Then you can use this prompt to audit what's missing and what connections you can make to help improve your verification. The third, for nontechnical work, you can use some of the skills that we made earlier in this video, the ask the board, the internal focus group, or you can use buildp partner.ai, the /expert advice, which has experts to get feedback on whatever you're working on. So, let's say you were creating a sales email, you could see what Alexi would think about it and if it approved or does not approve it.
In 2026, you need to become a gardener, not a sculptor. And the difference is whether the system catches its own mistakes before you have to.
Fix number four is that you have to stop treating everything like a nail. And this is inspired by Elon's five-step productivity framework. If you hand someone a hammer, everything starts looking like a nail. If you hand someone clawed, everything starts looking like an AI problem. But the reality is that most things you want to fix with Claude can be solved more effectively with a different lens before reaching for AI on any inefficiency run into run my five-step AI enhancement framework.
First, you need to question every requirement. Is this step in the process even necessary? Half of business processes exist because someone was too polite to delete them or you just never told someone it's not needed. Step two is delete the part or process. If you can kill it, kill it. Elon puts it bluntly. >> The most common error of a smart engineer is to optimize the thing that should not exist. >> If you aren't deleting so many processes that you have to add back 10% of them, that means you're not deleting enough. And trust me, this applies to both business owners and employees. I've been on both sides of it. When I was at JP Morgan, I deleted so much it just wasn't [music] needed. Step three, simplify or optimize. If something is absolutely needed, now you want to try and simplify the process. Remove steps, reduce friction, combine tasks. Step four is skill-driven AI augmentation. Tackle a task with AI as your assistant. And if the process repeats, create a clawed skill based on what you just did and then reuse that every time you run this process. Then each time you use a skill, work on enhancing it so that every iteration it gets better. Step five is skill-driven AI automation. In step four, you created a skill that you know works because you actively used it. Only after you've done that is it time to actually automate something. If you're automating something that you haven't already systematized manually, then you're just wasting your time. Personally, I don't trust myself to run this check in my head. So, I actually have an automation verification skill. If I'm debating building out a robust automation, this skill has to sign off on this decision based on everything it knows. Here's a prompt that you can use to build out a skill that you can just use as a sanity check whenever you're thinking about building out a crazy automation. If you just step right to four and five, you're going to be stuck working on things that you probably should have deleted or just replaced entirely. We've created a framework so that we can be a gardener, not a sculptor. And we understand not to use AI for everything.
But once it's actually time to use AI, one problem people run into is that they feel overwhelmed with all the new AI advancements. And before we get to what fixes this, which is something no other creator will talk about. If this is your first video of mine, welcome to the channel. If it's your second or more, here's our anti-slap agreement. The visuals, the testing, all of the things that are going on on screen. This is built entirely for humans, not for AI scrapers. So, all I ask part of this agreement is that you subscribe to this channel so that it can reach more people. Also, every video I give away a CloudMax subscription. And this week's winner is Maggra99. They're building a script to automate an old game. Honestly, hopefully that's Runescape. shout out my old character, Devils are back. And to enter the next giveaway, comment below with what you're building or a recent feature you built. And every new video you comment on will be an additional entry to this giveaway.
So, let's go to fix number five, which is about bridging the abstract to concrete. Every new a feature is new to everyone. So, no matter what, it's going to start off feeling very abstract. But each new feature or tool can be bridged to a concrete example for whatever you're working on. So, if you want to learn about claw skills, build a skill you can use. If you want to learn about clawed loops, build a loop you can use. The faster you go from abstract to concrete, the faster you'll build.
And so, how do you do this in practice? Well, first you want to do proof-based learning. Seeing results yourself is believing. So, concepts only stick when you've actually built something with them. So, you need a way to ground it to your work immediately for new concepts to stick. For example, when I wanted to learn about loops, I use this prompt. Look at my past sessions history and find tasks I've done multiple times where I'd benefit from a loop. Tell me how I could use this and why it's different than just calling skills. So when I ran this on the project I was working on, you can see the response on screen and at the bottom of it, it says, "So the loop does not replace the skill. It removes you as the trigger. Your three funnel digestion sessions across 8 days are exactly the symptom, a recurring obligation you carry in your head. The skill answers what work happens." The loop/scch schedule answers when and who remembers. This positions this new concept into the context of my project. And as a result, I'm able to understand how it applies to what I'm working on in minutes, not hours. The faster that you can bridge the abstract to the concrete, the faster you'll be able to build.
Fix number six, ship good, not perfect. One of my favorite quotes is from a French philosopher. Perfect is the enemy of good. The truth is, you can always make something better, but for most things, good is good enough, and any time you spent making it perfect is wasted. And I know what you're thinking. Doesn't this contradict what I said earlier about quality over quantity? The answer is no for two reasons. The first is that the best builders know what needs to be good versus perfect and they choose to build the minimum viable standard which saves them time and money. The second, when using AI, there's a concept of asmmptoic improvements. What this means is that AI is getting really good at getting you 80% of the way there. But no matter how many times you prompt it, it won't get you to that final 20%. So if you try and go from the 80%, the good to the great, the 100%, you may never actually get there by prompting AI. And there's a chance that you might actually see slight regression trying to make it perfect. So I'm not saying deliver AI slop. Good is not slop. What I am saying is follow the 8020 rule. 80% of your task should take 20% of the time. Lean on AI to get most of the way to that asmtote. 20% of your task should take the 80% of the time. That's where you want to focus your energy to bring good to great.
And I can't stress this enough for people watching who run their own business or want to run their own business. Peter Levelvels, a famous indie hacker who launched products, said that only four out of 70 plus projects that he's launched ever made money and grew. So about 95% of everything that he's done has failed. And this means in some cases, it's important to build things that are good and worry about making them perfect later. The reality is with all of this is that Claude is not the bottleneck, it's you. But if you apply these six fixes, you'll start building faster and more efficiently. Because the reality is, no matter what improvements come from AI, these bottlenecks will hold you back until you fix them.
Now, if you like this video, you'll love this video where I walk through exactly what you can paste into Claude to start building 10 times faster. The fixes I cover here, paired with the tactical advice on that video, puts you in the top 1% of all AI builders. I'll see you over there. pace.