Reverse Engineer Anything Does Way More Than You Think

0:00 Reverse Engineer Anything just went viral, and people are freaking out because it's a free tool that lets you rebuild almost any app in minutes. Now, I will show you how to use this, but

0:08 that's not the point of this video. Instead, I'm going to give you a step-by-step playbook to reverse engineer anything, including non-technical work. So, first, we'll

0:16 cover what REA is and use it to rebuild an app. Then, we'll apply the exact same playbook to reverse engineer four things: successful businesses, specific tools, tasks you want to automate, and

0:26 strategic thinking from industry experts. I guarantee whether you're technical or not, this will change how you use AI every single day. So what REA is and how it works. So REA, short for

0:36 reverse engineer Anything, is an open-source tool that anyone can use to rebuild a computer app. But it's actually nothing new. It's really just a combination of several pre-existing

0:45 tools that engineers have used for years. So Gidrah Hopper and IDA, which take apart compiled software, ChromeDev Tools, which show you what a website is doing, and app code reader, which walks

0:56 through an app. And these aren't new, right? Some of them have actually been around since the 1970s. So, what makes REA unique is it's a harness. It lets any person talk to all of these tools at

1:05 once without knowing how to use any of them. On screen, you can see a diagram directly from REA's own documentation, and it's essentially just a loop. First, you give your AI agent a target, for

1:15 example, an app. Then, REA inspects it and traces it with its analysis tools. Third, the agent then takes that information and then starts building the app. It then keeps looping until the app

1:24 is entirely recreated. So, there is a bigger story here about what this means for the future of software, and I'll get to that at the end of the video. But first, how do you actually use this

1:32 thing? So, first download RA by running this prompt. Then, you can pick the app that you actually want to recreate. On screen, I chose to reverse engineer Apple Notes. And after about 40 minutes,

1:41 here is the result. On the left is Apple's version, and on the right is the recreated one. They're pretty damn similar, and it is awesome. But again, that is not the point of this video.

1:50 You're likely a business owner or a highle operator watching this channel. How can you actually apply this to get value from it? Just ripping off a notes app is genuinely useless. So what should

2:00 you actually take from this? Well, REA has put a spotlight on something people have been doing forever. And that's figuring out exactly how something works, then reverse engineering it to

2:11 rebuild it. And in every field, not just technical ones, people have been doing this. So for example, writers, they learn by studying authors they admire. They take apart the sentences, borrow

2:20 techniques, and eventually they find out their own voice. Now, if you're not a writer, let's say in business, it's the same process. We borrow pieces of successful businesses and their

2:29 processes to make them our own. So, here is how you can actually do this yourself. Rea works in five steps. And we're going to borrow them for the rest of the video. First, pick what you're

2:38 studying. Second, gather the evidence. Third, design what you'll build from it. Four, rebuild and grade it against the original. Five, iterate yourself and make it yours. That's the process to

2:48 learn anything from winners and make yourself more productive. So, the first thing we're going to reverse engineer is businesses. There's a famous process Russell Brunson popularized called

2:57 funnel hacking. You walk through a competitor's funnel, which is how they turn strangers into customers, and study the exact steps they use. And we're going to systematize that using our

3:05 five-step REA framework so you can see what makes them so successful. So, first, pick the business you want to analyze. And a general rule of thumb for whatever business you're analyzing, the

3:15 bigger the player, the less you should study them. And that's because, let's say you're looking at Nike, their brand props up whatever work they're doing. So, you can't just redo what Nike is

3:24 doing because you don't have that brand authority. So, instead, pick the small competitor that just broke out and see what they're doing, right? So, use this prompt to analyze and identify your top

3:33 competitors. And know you don't need to run a business for this to be valuable. For example, if you're in marketing, see what your marketing competitors are doing. Second step, gather data you can

3:41 leverage. Run this prompt to search through a competitor's page for info. And you'll find that some websites block automated tools, so you won't be able to see their subpages in the funnel. And so

3:51 that's where you have a couple of options. One, you can have Claude walk through it in your browser. And the key here, whenever you do this, I would tell Claude to take screenshots of any key

3:59 steps. In my eyes, think of these screenshots as your evidence pile. They remove the interpretation gap because you can see what Claude saw instead of trusting its summary. But at this point,

4:09 you may still get blocked. And this is very slow. So if you want a faster, more effective way to do it that effectively look through each page and get through any blockers, I added a competitor

4:17 research tool to build partner.ai, which you can use entirely for free. And you can run this single command and it'll go through the funnel and do a deep analysis. Step three, design. In this

4:26 case, that means taking what you learned and applying it to your work. Here's a prompt that takes that information and gives you actionable takeaways for your business. Four is rebuild and grade.

4:36 take that list from the previous step and have Claude build what it can. Then evaluate the final product. And finally, step five is iterate. See what's working, see what didn't, see what other

4:45 people are doing over time. This is a continuous learning process. Now, as we go through each of these use cases, both REA as well as anything that we're going to cover, there are limitations. So,

4:54 neither this funnel hack nor REA can see the back end of how things work. So, in the business example, it can't see the customer support, [music] the email follow-ups, etc. And for REA and the

5:03 computer example, it can't see anything that it doesn't have direct access to. So for example, the backend or the database of the tool itself. But it doesn't have to be that extreme.

5:11 Sometimes the most valuable thing is a single tool that you recreate. But before we get to number two, building tools. Everything in this section was about studying how someone else turns

5:20 strangers into customers. What I can't show you is where those strangers come from the first place. Which brings us to today's video sponsor, Clay. So everyone on this channel is learning to build

5:28 with AI, but nobody really talks about who you're actually building it for. And finding customers is still the hardest part. Now, there are two ways to actually get these customers. First, you

5:36 can build an audience, which takes years, or go find the right people and reach out to them directly. And you can start doing that today. And that's what Clay is for, and that's what I've used

5:45 to scale my own business. So, what Clay does is it finds the businesses and people that need what you sell and tells you how to reach them directly. So, you describe the customer you want in plain

5:54 English, like find US companies with 10 to 100 people hiring an AI automation engineer right now, and get me the founder or head of operations at each one with their email. Clay will then

6:04 build a list, finds the decision makers, and researches each one so you know who's worth your time. A company that's hiring is telling you it's ready to buy, which makes your outreach feel smart,

6:14 not spammy. And Clay's new go to market agent is about to launch, and you can access it today directly through Clay's CLI in Claude. You tell it something like, "Build me an outbound campaign

6:22 that finds dental practices with five to 20 employees and writes a personalized email to each office manager." Then the Clay CLI will build the whole thing for you. It builds the go-to market

6:32 workflows for you, the outbound campaigns, the scoring flows, the sourcing data, finding lists of leads. It can do it all, which is pretty wild. And the thing that powers all this is

6:41 that Clay connects to over 250 data providers. And by pairing that data from all these providers together, Clay has access to the best data in the world, all in a single place. Now, on my

6:51 channel, I talk a lot about problems you should and shouldn't solve yourself. And setting up 250 data contracts is one that you should not solve. So, Clay is free to start, and no credit card is

7:00 needed. And if you sign up through the link in the description, my community gets $250 worth of free credits to get started, which you can get a lot done with. Some problems need one-time

7:09 solutions, and some problems are recurring. Finding new customers is a recurring problem and Clay provides a recurring solution. So that's about finding customers. The second thing that

7:18 we're going to reverse engineer is about building tools, not apps. On this channel, I've stress the importance of building for yourself, not for others. The reason is that if you build

7:27 something that you will use no matter what, you get value from it. And for me, this is actually where the whole build partner.ai road map comes from. I figure out what I need and I build it for me.

7:36 And then it usually works for people who also want to get the most out of Claude. And so when you're building for yourself, the key here is also that you don't want to try and do everything at

7:45 once. Every app or product that you see is really just a bundle of features and they're designed for the masses. So some features may be valuable to you, some features aren't. So when you're reverse

7:54 engineering existing apps, I would avoid rebuilding entire things, but instead focus on rebuilding specific features or tools. And before you actually go ahead and rebuild anything, you have to make

8:05 sure you understand why you're actually rebuilding it. It can be very tempting to fall into a dopamine building loop where it feels productive to recreate things, but in reality it's not actually

8:15 a needle mover. So, if you're just rebuilding something to rebuild it, don't do it. There has to be a concrete enhancement that you want to do that it can't currently do. And we'll cover that

8:24 a bit more in a sec. But once you've got a tool in mind, run it through our reverse engineering framework. First, pick one feature, not the entire app. Here is a prompt that can actually help

8:33 you do this. So, RA's own example does exactly this. It targets a feature, not the whole app. Their read me, their documentation says, "Understand how search works in the notes app. Show me

8:42 the evidence and build a similar feature for my project." And if Rea feels like too much, just use claw. Second step, gather information about the feature. This is where I love either screenshots

8:51 or recording of using the feature. I've taken thousands of screenshots of features that I like to inform product decisions and this is extremely effective. The third design and focus on

9:00 creating a minimum viable architecture depending on how you want to recreate the feature. You do not want to over complicate it. If it's one feature that stands on its own, have Claude make it

9:09 as an artifact. Don't spin up a whole app and worry about hosting and all the other stuff. Just let Claude do it for you in its artifact templates. The simplest solution is the best solution.

9:19 At this point, you should be ready for step four, which is rebuild the feature. Hand over the evidence and the design and let Claude RIP or use REA to rebuild it. Five, iterate from there and make it

9:29 yours. back to understanding why you'd be recreating a feature. In this reverse engineering scenario, there has to be some enhancement you're adding. Otherwise, it's useless. So, this step

9:38 is probably the most important. Now, building a tool is great, but the biggest time savings can come from something even simpler, and that's getting Claw to do a task for you. The

9:47 third category to reverse engineer, automate tasks. And we're going to record the best version and then hand it off. The goal here is to reverse engineer how you complete a task, then

9:56 take that and systematize Claude to do it for you. Here, we won't use REA in this step, but we're going to reuse the framework. And so, the simplest way to do this is actually just ask Claude to

10:05 review your past conversation history and flag any Claude skills you should create to systematize what you're doing. This will go through your conversation history and reverse engineer Claude

10:15 skills based on your past activities. And if you run that, you'll have repeatable Claude skills that you can use for task going forward. And you'll see in the prompt itself, an

10:23 optimization trick when you do mine your chats, stop asking, "What did I use AI for?" Instead, ask, "What did I have to send back?" Every time you send something back, that's a step worth

10:32 making into a skill. But if it got it right the first time, then you probably don't need a skill for it. But that is just level one of doing this. Level two is a lot more interesting. My favorite

10:41 way to reverse engineer a task is to record myself doing it. Within Claude, you can record your screen and Claude will identify ways to automate what you're doing on the recording. But just

10:50 clicking through the screens is only part of the process. You have to make sure to narrate while you work to give Claude additional insight into your thought process. That combines the

10:59 visual component with the audio component. Claude will see what you do and then hear what you say. And as you do that, make sure to mention what to do and why. You want to explain it like

11:08 you're teaching a human on an onboarding call. You would explain the logic behind your decisions. Then Cloud will take all this data and it's trained to suggest programmatic ways to do the task. It

11:17 won't just think about brute forcing it. It'll actually think about things like using APIs or MCP connections to automate this process. Then if you want to take it to the next level, you can

11:27 record a rubric, not just the task itself. So you can record yourself reviewing the finished output, saying what's good, what's wrong, and why. And then Claude can take that to iterate and

11:37 improve the skill over time. This is your way of teaching Claude the judgment, not just the clicks. So bring all of this through our framework. You want to first pick the task you do most.

11:45 Second, gather the info through the recording process. Third, lean on claw to design the automation. Four, rebuild it and then grade it by running it yourself. And then five, iterate until

11:54 it matches how you do it and provide feedback directly. At this point, we focused on getting AI to automate tasks, build out specific features, and study successful businesses. But the biggest

12:03 opportunity is to reverse engineer how industry experts think. Before we get to that, if this is your first video of mine, welcome to the channel. But if it's your second or more, this is our

12:11 anti-slap agreement. everything here, the hours of research, the hours of editing. This is built entirely by humans for humans. So, all that I ask as part of this agreement is you subscribe

12:20 to help this content reach more people. And if you already subscribed, you're a legend. But make sure to turn on those post notifications. And as a thank you to everyone who watches, I give away a

12:28 Claw Max subscription every video. So, this video's winner is YTP Pepper for building an MMO RPG. For this video, comment below with what you're building to enter the next giveaway. Let's get to

12:39 the next process we'll reverse engineer, which is strategic thinking. This is how you can reverse engineer how an expert thinks and works. Of everything in this video, this is the most important

12:48 section. So, I talked about earlier stealing like an artist. And this is the same idea at a more professional strategic level. And we're actually going to follow a Japanese martial arts

12:56 learning practice called shuhari. And here's how the practice breaks down. Shu, which means obey, ha, which means break, and ri which means move away. To explain all three of these steps,

13:07 there's a great example of this in the TV show The Bear. Richie spends a week working at a top fine dining restaurant where he starts by polishing forks. That shoe that's following someone else's

13:16 system exactly, even when you don't understand why yet. Then he starts to get what makes that restaurant works, right? He sees how the pieces fit together, what matters, and how those

13:25 practices could translate to a different environment. That's ha. And eventually, he brings the experience back to the bear where he and the team build something of their own. That's re taking

13:34 everything you've learned and developing an approach that works for your own goals. First you obey, then you break, and finally once you're ready, you move away. So how do you actually do this

13:43 with AI? Well, first pick who you want to learn from. And this can be people in your own circle or people that you respect from afar. Then we want to gather. We're going to gather

13:51 information from them. So we can download their brain directly to AI. For people in your circle, just collect information you know about them already. And so you can use clawed connectors to

14:00 pull in all their context, right? their calls and conversations, their research frameworks and written work, their previous reports, their emails, scripts, presentations, anything else that shows

14:09 how they think. And for people that you admire but you don't know, you'll need to work with public material. The simplest route is to use something like buildpartner.ai's expert advice, which

14:17 is pre-trained on top professionals. But if the person you want isn't there, you can clone them with build partner social tools, where you can run this prompt to search Instagram, Twitter, YouTube,

14:26 Substack, and more to gather their material. And if you don't know, build partner is my tool which is linked below if you do want to use it. Three, you then want to design a clawed skill using

14:35 that context. The goal isn't to just give AI a bunch of documents. It's to capture how this person thinks, what framework they use, and how they approach decisions. For someone in your

14:44 circle, you can build a skill around the material you've gathered. For someone who you admire from afar, you're building a model from their public body of work. Four, rebuild plus grade. This

14:53 is where you really put it to work. You could ask, "What would this person think about this situation? Give it a real problem to work through. See how it approaches the question, what it

15:00 prioritized, and what recommendations it makes. Then compare that what you know about the expert's actual thinking and work. Where does it get it right? Where does it miss something? What are the

15:09 assumptions it's making? How can you actually improve this? And now this is where you start understanding the underlying method, not just collecting someone else's opinions. [music] And

15:17 then finally, step five, iterate. Keep improving it as you find gaps, gather more context, correct assumptions, and refine the skill. Over time, you're moving through the same shoa

15:26 progression. first learning someone else's method, then understanding how to adapt it, and eventually developing your own. That's the real opportunity here. You're not just creating an AI version

15:34 of someone else's expertise. You're using it as a way to build out your own. Now, let's zoom out for a second and bring it all together. Rea is a tool that rebuilds apps. But the real lesson

15:43 is the five-step process behind it. Pick, gather, design, rebuild, plus grade, and then iterate. It's the same loop every time, but you want to copy the process, not the property. And if

15:53 you're still hesitant, Paul Graham put it very well. There's no better way to learn how something works than by trying to reproduce it. Nor does copying necessarily make your work unoriginal.

16:02 Now, I will say this tool rea is scary because it calls into question the value of products in our economy. So, what do I think of the future of software? If anyone can rebuild how an app works in

16:12 minutes, the code stops being the mode. So, what is actually left that's hard to copy? Let's go back to what REA can't see. That's the backend, the relationships, the trust, and the

16:21 judgment about what to build and why. When the cost of building a feature drops towards zero, the scarce thing isn't the software. It's about knowing what you need and having the taste to

16:28 know when it's good. The winners won't be the people who build the most. They'll be the ones who know what's worth building for their specific customer. And building on that, a

16:36 mindset I say to my team, we build things. You can't just vibe code. So, what are the things that make you special? So, if it's Uber, it's not the app. It's the driver. If it's Airbnb,

16:44 it's not the app. It's the hosts and the guests. There's plenty of things that make you valuable that REA just can't copy and paste. Now, I touched on it earlier, but one problem you'll run into

16:53 is this non-productive building dopamine loop where you continuously build things that have no real value. And the most effective AI builders know what to build and what not to build. And to get a

17:03 better understanding of this, go check out this video where I share the five questions you can ask yourself to build 10 times faster. If you're new here, my name's Austin Maresy. Welcome to the

17:12 channel and I'll see you in the next video. Pace.