Matt Pocock: The Most Valuable Skill to Learn in the Age of AI

How do you stand out in a world where anyone can where anyone can where anyone can create things so create things so create things so quickly? quickly? quickly? I think the truly I think the truly I think the truly successful people are the ones successful people are the ones successful people are the ones who are the best at who are the best at who are the best at communicating, communicating, communicating, right? So, in my opinion, the right? So, in my opinion, the right? So, in my opinion, the best way best way best way to get to get to get great work from agents is to great work from agents is to great work from agents is to tell them what's important to you tell them what's important to you tell them what's important to you . . . So, the main skill So, the main skill is the ability to just is the ability to just is the ability to just put it all out there from put it all out there from put it all out there from your head. your head. your head. This skill is about This skill is about This skill is about being aware of being aware of being aware of unknown unknowns, unknown unknowns, unknown unknowns, right? It could right? It could right? It could be security, be security, be security, deployment method, or deployment method, or deployment method, or just things just things just things you haven't thought about. And if you haven't thought about. And if you haven't thought about. And if you're not careful you're not careful you're not careful about this, the agent will about this, the agent will about this, the agent will just start just start just start making up nonsense. making up nonsense. making up nonsense. What great What great What great opportunities do you see opportunities do you see opportunities do you see to to to help start your help start your help start your career in the right career in the right career in the right direction? direction? direction? Using agents Using agents , developing processes , developing processes , developing processes for intellectual for intellectual for intellectual work, and then work, and then work, and then implementing them in implementing them in implementing them in organizations. After all, organizations. After all, organizations. After all, every organization every organization every organization does things a little does things a little differently. I think this is a differently. I think this is a differently. I think this is a huge huge huge opportunity. opportunity. opportunity. I would like to hear I would like to hear I would like to hear some story from the " some story from the " some story from the " battlefield." battlefield." battlefield." I tried I tried I tried to do to do to do spectral spectral spectral development, I guess. And I development, I guess. And I development, I guess. And I just saw the code just saw the code just saw the code get worse get worse get worse and worse. And the agent and worse. And the agent and worse. And the agent produced worse produced worse produced worse and worse results. I and worse results. I and worse results. I tried this for months, tried this for months, tried this for months, over and over again over and over again over and over again . And it didn't . And it didn't . And it didn't work. So I got it. work. So I got it. Okay, Matt. I'm very Okay, Matt. I'm very happy to be sitting here with happy to be sitting here with happy to be sitting here with you today. You know you today. You know , of course, I often , of course, I often , of course, I often use your use your use your skills. And I'm skills. And I'm skills. And I'm really curious that you really curious that you really curious that you learned to code on your learned to code on your learned to code on your own. And when we own. And when we own. And when we first started first started first started talking, before the talking, before the talking, before the recording, you said that recording, you said that recording, you said that learning code is much learning code is much learning code is much easier than people easier than people easier than people say. So, you say. So, you say. So, you learned learned learned to program on your to program on your to program on your own. You did own. You did own. You did this yourself for about this yourself for about this yourself for about 10 years. And then we 10 years. And then we 10 years. And then we entered this era where there entered this era where there entered this era where there were agents were agents were agents for writing code and for writing code and for writing code and so on. So I'm really so on. So I'm really so on. So I'm really interested to hear your

interested to hear your interested to hear your thoughts on all of this thoughts on all of this thoughts on all of this today. But I would today. But I would today. But I would like to like to like to ask right away: what ask right away: what ask right away: what was the transition like for you was the transition like for you was the transition like for you when you learned all the when you learned all the when you learned all the basics of programming basics of programming basics of programming and wrote code yourself and wrote code yourself , and then this , and then this , and then this term "vibe coding" appeared and term "vibe coding" appeared and term "vibe coding" appeared and people started people started people started creating everything in creating everything in creating everything in one day? What was one day? What was one day? What was this transition like this transition like this transition like for you? Well for you? Well for you? Well , thank you. Thank you for , thank you. Thank you for , thank you. Thank you for using using using my skills as well. I my skills as well. I my skills as well. I appreciate it. That's very appreciate it. That's very appreciate it. That's very kind of you. kind of you. kind of you. Um, I really started out as a Um, I really started out as a skeptic for a long time, skeptic for a long time, you know? I had you know? I had you know? I had about 10 years of about 10 years of programming experience, programming experience, first as an amateur and first as an amateur and first as an amateur and then about 8 years then about 8 years then about 8 years professionally, before the professionally, before the professionally, before the advent of "vibe coding" advent of "vibe coding" advent of "vibe coding" and before everything and before everything and before everything became really became really became really high-quality. I think the high-quality. I think the high-quality. I think the release of Opus 4.5 and release of Opus 4.5 and release of Opus 4.5 and the creation of Open Claw the creation of Open Claw the creation of Open Claw around December of around December of around December of last year, 2025, last year, 2025, last year, 2025, was the same was the same was the same moment, the threshold, moment, the threshold, moment, the threshold, when I said to myself, “ when I said to myself, “ Okay, I’m never going Okay, I’m never going Okay, I’m never going to write another to write another to write another line of code by hand again line of code by hand again .” So this was the .” So this was the .” So this was the moment when agents moment when agents moment when agents became really, became really, really good really good . And the whole year before that . And the whole year before that . And the whole year before that I was thinking, “Okay, yes, I I was thinking, “Okay, yes, I I was thinking, “Okay, yes, I see the benefits. I see see the benefits. I see see the benefits. I see everyone is really excited about everyone is really excited about everyone is really excited about it, but it has it, but it has it, but it has n't quite n't quite n't quite worked for me yet." And as soon as worked for me yet." And as soon as worked for me yet." And as soon as I I don't know. What I I don't know. What I I don't know. What was always interesting to me was always interesting to me was always interesting to me in my work was in my work was in my work was helping helping helping my colleagues become my colleagues become my colleagues become better, you know?

better, you know? better, you know? Trying to create Trying to create Trying to create tools that tools that tools that helped colleagues helped colleagues helped colleagues release release release products faster, trying to products faster, trying to products faster, trying to communicate with communicate with communicate with other developers other developers other developers and teach them and teach them and teach them to do their to do their to do their jobs better, even if jobs better, even if jobs better, even if they were perhaps they were perhaps they were perhaps older than me. So I was older than me. So I was older than me. So I was always interested in always interested in always interested in the environment in which the environment in which the environment in which my colleagues worked my colleagues worked my colleagues worked and the environment in and the environment in and the environment in which I myself worked. which I myself worked. which I myself worked. And when agents came along And when agents came along And when agents came along who were who were who were really effective, really effective, really effective, I started thinking, " I started thinking, " Great, I can Great, I can Great, I can use the skills I've use the skills I've use the skills I've learned throughout my learned throughout my learned throughout my career and my career and my career and my interest in this field interest in this field to help these to help these to help these agents become agents become agents become much better and much better and much better and much faster." And much faster." And much faster." And until Opus 4.5 came along, it didn't until Opus 4.5 came along, it didn't until Opus 4.5 came along, it didn't work very work very work very well. And then, well. And then, well. And then, last December, last December, last December, it worked. And I it worked. And I it worked. And I realized, "Wow, okay, realized, "Wow, okay, realized, "Wow, okay, I can really I can really I can really apply my apply my apply my skills here and skills here and skills here and create some pretty create some pretty create some pretty cool stuff." And so far cool stuff." And so far cool stuff." And so far it's working very well. it's working very well. it's working very well. This is super cool. You know This is super cool. You know , one of the things I've , one of the things I've , one of the things I've taken away from what you've taken away from what you've taken away from what you've been posting online is that been posting online is that been posting online is that you think the you think the you think the fundamentals and principles of fundamentals and principles of fundamentals and principles of software engineering are software engineering are software engineering are even more important now even more important now as AI is as AI is being implemented into all being implemented into all being implemented into all these development processes.

these development processes. But at the same time, you don't But at the same time, you don't necessarily think necessarily think necessarily think that a person has to be that a person has to be that a person has to be super technical or super technical or super technical or have a lot of have a lot of have a lot of programming experience programming experience to use to use to use these tools. these tools. these tools. So what does this look like So what does this look like So what does this look like for someone who is not a for someone who is not a techie? Because I techie? Because I know that your know that your know that your audience is mostly audience is mostly audience is mostly tech-savvy. tech-savvy. tech-savvy. What does this look like for What does this look like for What does this look like for someone who is not a someone who is not a someone who is not a technical technical technical expert? How can they expert? How can they use the use the fundamentals and principles of fundamentals and principles of fundamentals and principles of software engineering software engineering to build to build to build products correctly? products correctly? products correctly? That's right. So, in my That's right. So, in my That's right. So, in my opinion, the best opinion, the best opinion, the best way to get way to get way to get the most the most the most out of agents is to tell out of agents is to tell out of agents is to tell them what's important to you them what's important to you , right? Because agents , right? Because agents , right? Because agents tend, like humans, tend, like humans, tend, like humans, to underestimate the extent to which to underestimate the extent to which to underestimate the extent to which humans humans humans underestimate the underestimate the underestimate the communication barrier between you, the communication barrier between you, the communication barrier between you, the human, and the agent, human, and the agent, human, and the agent, right? And no matter right? And no matter right? And no matter how how how smart this model is, there will smart this model is, there will smart this model is, there will always be always be always be this chasm between you. And the thing is, the this chasm between you. And the thing is, the this chasm between you. And the thing is, the agent doesn't know agent doesn't know agent doesn't know exactly what you want, exactly what you want, exactly what you want, he doesn't know what's he doesn't know what's he doesn't know what's important to you, so you important to you, so you important to you, so you have to have to have to repeat it over repeat it over repeat it over and over again every time. So I think and over again every time. So I think you'll get you'll get you'll get better results from better results from better results from agents if the agents if the agents if the environment environment environment they're working in is they're working in is they're working in is suitable for them, suitable for them, suitable for them, right? If they're right? If they're right? If they're working with some working with some working with some horrible, horrible, horrible, hastily hastily hastily written written written codebase that codebase that codebase that no one cares about and no one cares about and no one cares about and only cares about the results, only cares about the results, only cares about the results, then you're going to get then you're going to get then you're going to get garbage out of it, garbage out of it, garbage out of it, right? The agent right? The agent right? The agent will simply produce junk will simply produce junk will simply produce junk because it is because it is because it is in a junk in a junk in a junk environment. So you will have to environment. So you will have to environment. So you will have to start

start start taking care of this taking care of this taking care of this environment. And you environment. And you environment. And you can start to care about that can start to care about that can start to care about that by thinking, by thinking, by thinking, let's say, if I were let's say, if I were let's say, if I were running a huge running a huge development organization and I had 100 development organization and I had 100 people, people, human programmers—I know, it human programmers—I know, it human programmers—I know, it sounds weird—to sounds weird—to sounds weird—to actually actually actually do my do my do my work, right? Or work, right? Or work, right? Or that they produce that they produce that they produce results. How would I results. How would I results. How would I raise the bar so that raise the bar so that raise the bar so that the worst programmer the worst programmer I have I have I have access to can do a access to can do a access to can do a better job? What strategies would better job? What strategies would better job? What strategies would I use? And I use? And I use? And it turns out people it turns out people it turns out people have been thinking about this for 20 have been thinking about this for 20 have been thinking about this for 20 years, right? And years, right? And entire books have been written about all these concepts. entire books have been written about all these concepts. And it turns out, if And it turns out, if And it turns out, if you read these books, it you read these books, it you read these books, it seems as if they seems as if they seems as if they were written were written were written last week. They last week. They last week. They seem so seem so seem so relevant now relevant now relevant now because they are about because they are about because they are about how to create how to create how to create quality tests. How to quality tests. How to quality tests. How to add checks add checks add checks to make your codebase get to make your codebase get better over time, to raise better over time, to raise the bar, stop entire the bar, stop entire the bar, stop entire classes of bugs, and classes of bugs, and classes of bugs, and prevent mistakes. prevent mistakes. prevent mistakes. And that's exactly what everyone is And that's exactly what everyone is And that's exactly what everyone is trying to do now trying to do now trying to do now . So I think the . So I think the . So I think the first step to that is to start first step to that is to start caring about it, caring about it, caring about it, to strive to learn about it, and to to strive to learn about it, and to to strive to learn about it, and to be interested in it, because be interested in it, because be interested in it, because agents are also quite agents are also quite agents are also quite good at teaching good at teaching good at teaching you. So, if you're you. So, if you're you. So, if you're interested in this, if interested in this, if interested in this, if you ask the right you ask the right you ask the right questions, you know, questions, you know, questions, you know, agents will be able to agents will be able to agents will be able to understand that understand that understand that this is important to you and will start this is important to you and will start this is important to you and will start to guide you in to guide you in to guide you in that direction.

that direction. that direction. M-hm. This is extremely M-hm. This is extremely M-hm. This is extremely interesting, because I interesting, because I interesting, because I can certainly say that 80% of can certainly say that 80% of what I have learned what I have learned what I have learned in the last year, I have in the last year, I have in the last year, I have mastered thanks to mastered thanks to mastered thanks to my agents. And I my agents. And I my agents. And I know you have the know you have the know you have the skill {slash} teach skill {slash} teach . And you know, . And you know, . And you know, I'm wondering how you I'm wondering how you I'm wondering how you feel about trusting what feel about trusting what feel about trusting what they they they teach you? teach you? You shouldn't You shouldn't trust, right? You trust, right? You trust, right? You shouldn't. I think it's shouldn't. I think it's shouldn't. I think it's always a good idea to be always a good idea to be always a good idea to be skeptical of everything skeptical of everything skeptical of everything you're told. I guess I'm a you're told. I guess I'm a born skeptic. It born skeptic. It took me a long took me a long took me a long time to time to time to get interested in AI because I get interested in AI because I get interested in AI because I thought it was thought it was thought it was nonsense. It's like the nonsense. It's like the nonsense. It's like the new new new web3 cryptocurrency bubble, you know, web3 cryptocurrency bubble, you know, web3 cryptocurrency bubble, you know, complete nonsense and all complete nonsense and all complete nonsense and all that. Am I allowed to that. Am I allowed to that. Am I allowed to swear here? I'm not swear here? I'm not swear here? I'm not sure. I tend to sure. I tend to sure. I tend to swear. Good. Okay, swear. Good. Okay, swear. Good. Okay, great. I often swear great. I often swear great. I often swear during podcasts. I don't during podcasts. I don't during podcasts. I don't know. It helps know. It helps know. It helps me relax. me relax. me relax. So, I was So, I was So, I was very skeptical for a long time. And I very skeptical for a long time. And I very skeptical for a long time. And I remain skeptical remain skeptical until I can until I can until I can read the read the read the original source, okay? original source, okay? So when I So when I developed the developed the developed the learning skill, it was learning skill, it was learning skill, it was basically about basically about basically about finding reliable, finding reliable, finding reliable, high-quality high-quality high-quality primary sources and then primary sources and then primary sources and then explaining them to explaining them to explaining them to people with citations, people with citations, people with citations, references, and references, and references, and directions to where to directions to where to directions to where to find them. Over time, the find them. Over time, the find them. Over time, the AI ​​agents' hallucinations AI ​​agents' hallucinations have significantly decreased. You know, have significantly decreased. You know, they don't they don't they don't make up information that often anymore make up information that often anymore . They still do . They still do . They still do it. Fable still does it. Fable still does it. Fable still does this. They still this. They still this. They still do it, but do it, but do it, but much less often. And I much less often. And I much less often. And I realized that realized that realized that most often most often most often I use these I use these I use these training sessions only to training sessions only to introduce me to the introduce me to the topic, and then I

topic, and then I topic, and then I go read go read go read the primary source myself, because the primary source myself, because the primary source myself, because it contains the it contains the information I need. It's just that information I need. It's just that I may have I may have I may have lacked lacked lacked the terminology to the terminology to the terminology to understand him before the understand him before the understand him before the training training training session began. session began. session began. So guys, really So guys, really So guys, really quickly, a huge quickly, a huge quickly, a huge thank you to Hostinger for thank you to Hostinger for thank you to Hostinger for sponsoring this sponsoring this sponsoring this part of the video. I know part of the video. I know part of the video. I know many of you many of you many of you are learning to build are learning to build are learning to build AI, and at some AI, and at some AI, and at some point your development point your development point your development needs to live somewhere other than just needs to live somewhere other than just needs to live somewhere other than just your your your local laptop. local laptop. local laptop. Because once Because once others start using them, others start using them, you'll need a you'll need a you'll need a server that's up and running server that's up and running server that's up and running all the time and allows you all the time and allows you all the time and allows you to make changes without to make changes without to make changes without breaking what's already breaking what's already breaking what's already working. So, Hostinger VPS working. So, Hostinger VPS is your own is your own is your own server with dedicated server with dedicated server with dedicated CPU, CPU, CPU, RAM, RAM, RAM, and full root and full root and full root access. This makes access. This makes access. This makes setup the easy setup the easy setup the easy part of the job, so part of the job, so part of the job, so you can spend you can spend you can spend your time on development your time on development your time on development and not on technical and not on technical and not on technical details, because I know details, because I know details, because I know many of us don't many of us don't many of us don't have DevOps experience. have DevOps experience. have DevOps experience. So, I would choose the So, I would choose the So, I would choose the KVM 2 plan. For this price you KVM 2 plan. For this price you KVM 2 plan. For this price you get two vCPUs, 8 GB of get two vCPUs, 8 GB of get two vCPUs, 8 GB of RAM, RAM, RAM, and 100 GB of storage, which is and 100 GB of storage, which is and 100 GB of storage, which is quite enough for a quite enough for a quite enough for a first project. And you first project. And you first project. And you can always can always can always upgrade your plan upgrade your plan upgrade your plan later if the later if the later if the need arises. You can then need arises. You can then need arises. You can then select a Cloud Code template that select a Cloud Code template that select a Cloud Code template that already has Cloud Code already has Cloud Code already has Cloud Code installed. So you installed. So you installed. So you can start can start can start developing right developing right developing right after logging in after logging in . And to protect . And to protect . And to protect your work, you can your work, you can your work, you can take snapshots take snapshots take snapshots of the server before of the server before any changes and any changes and any changes and roll back roll back roll back in case of failures thanks to in case of failures thanks to in case of failures thanks to weekly weekly weekly backups. And if you backups. And if you backups. And if you get stuck, the get stuck, the get stuck, the Hostinger agent in the web console Hostinger agent in the web console Hostinger agent in the web console will help with will help with will help with server tasks, server tasks, server tasks, errors, and errors, and errors, and deployments, simply deployments, simply deployments, simply by understanding your by understanding your by understanding your normal language. So, normal language. So, normal language. So, you can purchase KVM you can purchase KVM you can purchase KVM 2 at hostinger.com/nateherkvps.

2 at hostinger.com/nateherkvps. 2 at hostinger.com/nateherkvps. The link is in the description, and The link is in the description, and The link is in the description, and you can you can you can use the code nateherk use the code nateherk use the code nateherk at at at checkout to checkout to checkout to get an additional get an additional get an additional 10% off any 10% off any 10% off any annual plan above the annual plan above the annual plan above the current price. current price. current price. Thanks again to Hostinger for Thanks again to Hostinger for Thanks again to Hostinger for sponsoring this sponsoring this sponsoring this part of the video, and part of the video, and part of the video, and now let's get back now let's get back now let's get back to business. Got it. to business. Got it. to business. Got it. OK. So, whatever you OK. So, whatever you OK. So, whatever you say, you see a say, you see a say, you see a lot of different people lot of different people lot of different people coming into your coming into your coming into your ecosystem, they're ecosystem, they're ecosystem, they're learning to master learning to master learning to master things with AI, things with AI, things with AI, like building like building like building products or products or products or automation, and automation, and automation, and they probably have a lot of they probably have a lot of they probably have a lot of different motivations. What different motivations. What different motivations. What opportunities do you opportunities do you opportunities do you think are emerging think are emerging think are emerging for people who want to for people who want to for people who want to see, see, see, maybe they're a little maybe they're a little maybe they're a little skeptical, like you, skeptical, like you, skeptical, like you, about this whole AI thing, about this whole AI thing, about this whole AI thing, but now we're seeing but now we're seeing but now we're seeing it, right? This is it, right? This is it, right? This is not going away. What are some not going away. What are some not going away. What are some great opportunities that great opportunities that great opportunities that you think are worth you think are worth you think are worth considering to considering to considering to help, you know, help, you know, help, you know, start your career in the start your career in the start your career in the right direction right direction ? ? Well, I mean, in Well, I mean, in any period of any period of any period of massive change, there are massive change, there are massive change, there are always always always opportunities, right? opportunities, right? opportunities, right? As things As things As things change, people change, people change, people will adopt will adopt will adopt technology technology technology more slowly. Um, as for more slowly. Um, as for more slowly. Um, as for me, I see a lot of me, I see a lot of learning opportunities, right?

learning opportunities, right? People want to study People want to study People want to study it because it is very it because it is very it because it is very useful and very useful and very useful and very powerful. And every powerful. And every powerful. And every organization I organization I organization I talk to encounters talk to encounters talk to encounters resistance resistance resistance when trying to when trying to when trying to implement these things. And implement these things. And implement these things. And let me let me let me clarify: I do clarify: I do clarify: I do believe that there believe that there believe that there are parts of the AI ​​world that are are parts of the AI ​​world that are are parts of the AI ​​world that are unethical or have unethical or have unethical or have questionable ethics. And I questionable ethics. And I questionable ethics. And I think there are quite think there are quite think there are quite valid thoughts that valid thoughts that valid thoughts that it's a little strange it's a little strange it's a little strange that Anthropic just that Anthropic just that Anthropic just stole a bunch of stole a bunch of stole a bunch of novels and then novels and then novels and then somehow paid somehow paid somehow paid for them later. for them later. You know, it's surprising that so You know, it's surprising that so much AI much AI much AI would put a would put a would put a strain on the strain on the strain on the energy energy energy ecosystem. But you're ecosystem. But you're ecosystem. But you're right, he's already right, he's already right, he's already here. It will not here. It will not here. It will not disappear anywhere, because there is an disappear anywhere, because there is an disappear anywhere, because there is an economic incentive economic incentive economic incentive for its development. for its development. So, as for So, as for the opportunities, they are the opportunities, they are the opportunities, they are everywhere you look: everywhere you look: everywhere you look: this includes creating various this includes creating various this includes creating various programs that programs that programs that embed AI, and embed AI, and embed AI, and coming to companies and coming to companies and coming to companies and saying: "I know how saying: "I know how saying: "I know how to use AI, and to use AI, and to use AI, and I can help I can help I can help improve and improve and improve and develop your develop your develop your systems." It seems like this is the systems." It seems like this is the buzzword on everyone's lips, and I buzzword on everyone's lips, and I believe that the opportunities are there believe that the opportunities are there believe that the opportunities are there if you look for them. I if you look for them. I if you look for them. I completely agree. So completely agree. So completely agree. So let's look at a let's look at a let's look at a practical example.

practical example. practical example. Let's say someone Let's say someone Let's say someone is currently working full- is currently working full- is currently working full- time and is not a time and is not a time and is not a technical professional. technical professional. technical professional. Let's say, maybe this Let's say, maybe this Let's say, maybe this person is just a person is just a person is just a salesperson. So, not salesperson. So, not salesperson. So, not a techie, working a techie, working a techie, working full-time, but full-time, but full-time, but wanting to get into wanting to get into wanting to get into AI: should I AI: should I AI: should I learn to build learn to build learn to build it myself and it myself and it myself and open a open a open a consulting consulting consulting business, or is it better business, or is it better business, or is it better to use AI to use AI to use AI at work, get at work, get at work, get promoted, and promoted, and promoted, and try to get into a try to get into a try to get into a management position? How would management position? How would management position? How would you evaluate these you evaluate these you evaluate these different paths and different paths and different paths and try to understand try to understand which one is which one is which one is right for you? right for you? I would say this: I I would say this: I look at it quite look at it quite look at it quite narrowly, namely—what narrowly, namely—what narrowly, namely—what skills will be skills will be skills will be useful to me in the useful to me in the useful to me in the future? And if future? And if future? And if you have the skills you have the skills you have the skills I'm about to I'm about to I'm about to share, you share, you share, you can find your can find your can find your way to something way to something way to something worthwhile, whether it's worthwhile, whether it's worthwhile, whether it's starting your own starting your own starting your own business or moving up the business or moving up the business or moving up the career ladder within career ladder within career ladder within your organization. And your organization. And your organization. And the skills that I find the skills that I find the skills that I find useful are...people useful are...people useful are...people think agents don't think agents don't think agents don't need training, need training, need training, right? They right? They right? They think that just think that just think that just talking to talking to talking to him is enough and getting him is enough and getting him is enough and getting results. But what results. But what results. But what I found out is that this is I found out is that this is I found out is that this is not true at all. And as not true at all. And as not true at all. And as models models models improve, the improve, the improve, the learning curve doesn't learning curve doesn't learning curve doesn't necessarily become necessarily become necessarily become flatter. The flatter. The flatter. The learning curve is still there learning curve is still there , it just might , it just might , it just might be a little be a little be a little hidden from hidden from hidden from most people. So most people. So , the concept: what , the concept: what , the concept: what can you trust when can you trust when can you trust when working with an agent? How do working with an agent? How do working with an agent? How do you organize you organize you organize the work of agents? How do the work of agents? How do the work of agents? How do you get agents to, uh, you get agents to, uh, you get agents to, uh, behave exactly the way

behave exactly the way behave exactly the way you want them to? How you want them to? How you want them to? How to control to control to control the information that the information that the information that agents receive? How do you agents receive? How do you agents receive? How do you develop develop develop the information you the information you the information you want to provide to want to provide to want to provide to agents? What, uh, agents? What, uh, agents? What, uh, forms and forms and forms and data structures should it be in? data structures should it be in? data structures should it be in? How do you then turn How do you then turn How do you then turn these agents into, uh, these agents into, uh, these agents into, uh, reusable processes reusable processes ? Or how do you take your ? Or how do you take your ? Or how do you take your processes and processes and processes and turn them into something turn them into something turn them into something agents can agents can agents can do? do? do? For example, I For example, I For example, I use agents use agents use agents for code, quite for code, quite for code, quite obviously. Agents are really obviously. Agents are really obviously. Agents are really good at good at good at code because it's code because it's code because it's easy to test, easy to test, easy to test, you know? Just like you know? Just like you know? Just like you see you see you see agents solving agents solving agents solving math problems, math problems, math problems, because math is because math is because math is very simple, very simple, very simple, sorry, sorry, sorry, math is very math is very math is very easy...well, she easy...well, she easy...well, she got it, right? You got it, right? You got it, right? You can create can create can create proofs that proofs that proofs that are generated over time are generated over time are generated over time and solve these and solve these and solve these incredible incredible incredible mathematical problems. mathematical problems. But real work is But real work is much more difficult much more difficult much more difficult than coding for than coding for than coding for agents because it is agents because it is agents because it is harder to verify. This is harder to verify. This is harder to verify. This is why I had the greatest why I had the greatest why I had the greatest difficulty difficulty difficulty trying to trying to trying to implement agents into implement agents into implement agents into my daily my daily my daily work; It took a work; It took a work; It took a long time and long time and long time and required required required trying out a lot of trying out a lot of trying out a lot of processes until I processes until I processes until I finally found something that finally found something that worked for me, and worked for me, and worked for me, and I'm slowly I'm slowly I'm slowly refining it over refining it over refining it over time. So I think that's where the time. So I think that's where the big opportunities lie. It's big opportunities lie. It's really about really about really about using agents using agents , creating processes , creating processes , creating processes for intellectual for intellectual for intellectual work, and then work, and then work, and then implementing them in implementing them in implementing them in organizations, because organizations, because organizations, because every organization every organization every organization works a little differently, works a little differently, works a little differently, and there are a lot of and there are a lot of and there are a lot of opportunities opportunities opportunities to tweak that.

to tweak that. to tweak that. Mhm. Mhm. Mhm. And this is a very difficult And this is a very difficult And this is a very difficult skill. Even I, skill. Even I, skill. Even I, having worked with having worked with having worked with agents for about a agents for about a agents for about a year and a half, still year and a half, still year and a half, still find it very find it very find it very challenging, and I think challenging, and I think challenging, and I think it's a huge it's a huge it's a huge opportunity. opportunity. Mhm. So, you said Mhm. So, you said that that that fundamentals are important. fundamentals are important. You also mentioned You also mentioned explaining the context of explaining the context of explaining the context of why you want why you want why you want something done. something done. something done. So it's kind of like So it's kind of like So it's kind of like prompting and prompting and intent-driven development. And how do these things intent-driven development. And how do these things fit together fit together fit together ? Where it fails is ? Where it fails is ? Where it fails is when a non-technical when a non-technical when a non-technical person might not know person might not know person might not know how to set up the how to set up the how to set up the right right right codebase or how to, you know, codebase or how to, you know, codebase or how to, you know, create API keys with the create API keys with the create API keys with the right right right permissions, or whatever the permissions, or whatever the permissions, or whatever the problem might be. problem might be. How can such a person How can such a person act with act with act with intent and still intent and still intent and still ensure that everything is ensure that everything is ensure that everything is done done done properly? I'm not properly? I'm not properly? I'm not sure if sure if sure if you have any stories from you have any stories from you have any stories from experience, and I'm experience, and I'm experience, and I'm sure this is where sure this is where your your "vibe coding "vibe coding " approach comes into play, uh, as a kind of " approach comes into play, uh, as a kind of " approach comes into play, uh, as a kind of transition from the old one. transition from the old one.

But I'd like to hear But I'd like to hear if you've seen any if you've seen any if you've seen any specific examples specific examples specific examples where this doesn't work where this doesn't work , and how someone might , and how someone might , and how someone might think about...I guess I'm think about...I guess I'm think about...I guess I'm trying to say what I trying to say what I trying to say what I 've seen myself: you 've seen myself: you 've seen myself: you can have can have can have enough enough enough awareness to awareness to awareness to get your agent to get your agent to get your agent to think about a problem, think about a problem, think about a problem, even if you even if you even if you don't know exactly how don't know exactly how don't know exactly how to solve it yourself. But if you to solve it yourself. But if you to solve it yourself. But if you can identify that can identify that can identify that this is something that needs to be this is something that needs to be this is something that needs to be addressed, that is often addressed, that is often addressed, that is often enough enough enough to make significant to make significant to make significant progress. So, I guess progress. So, I guess my point is, my point is, my point is, how do you how do you how do you think the average think the average think the average person can discover person can discover person can discover these kinds of opportunities? these kinds of opportunities? So, you're talking So, you're talking about a regular person about a regular person about a regular person versus a programmer, versus a programmer, versus a programmer, right? But everything you right? But everything you right? But everything you say sounds very familiar to me as a say sounds very familiar to me as a say sounds very familiar to me as a developer developer developer , doesn't it? There , doesn't it? There , doesn't it? There is no is no is no distinction here. distinction here. There is a great There is a great quote in the book The Pragmatic Programmer: "Nobody quote in the book The Pragmatic Programmer: "Nobody quote in the book The Pragmatic Programmer: "Nobody knows exactly what they want, knows exactly what they want, knows exactly what they want, right?" Nobody knows right?" Nobody knows what they want. what they want. what they want. Everyone has these Everyone has these Everyone has these "unknown unknowns" in their minds. "unknown unknowns" in their minds. "unknown unknowns" in their minds. Every time they Every time they Every time they clarify something, clarify something, clarify something, every time they every time they every time they say, let's say, to say, let's say, to say, let's say, to a developer—when I'm a a developer—when I'm a a developer—when I'm a regular person regular person regular person talking to talking to talking to a developer, which is like a a developer, which is like a a developer, which is like a regular person regular person regular person talking to an talking to an talking to an agent—there's this agent—there's this agent—there's this communication communication communication barrier, right? And barrier, right? And barrier, right? And the developer will listen to the developer will listen to the developer will listen to everything and try to everything and try to everything and try to figure out what is important to me figure out what is important to me figure out what is important to me . So one of . So one of . So one of my skills, in particular my skills, in particular , which is one of the , which is one of the , which is one of the most most most used, used, used, is called "question is called "question is called "question me." And the gist of it is me." And the gist of it is me." And the gist of it is that I'm essentially that I'm essentially that I'm essentially ordering the agent to ordering the agent to ordering the agent to relentlessly interrogate relentlessly interrogate relentlessly interrogate me about the thing we're me about the thing we're me about the thing we're trying trying trying to build. And the reason to build. And the reason it's so it's so it's so effective is because

effective is because effective is because it gets it gets it gets more context from more context from more context from you about what's important to you you about what's important to you you about what's important to you and what you're and what you're and what you're actually actually actually trying to create trying to create , and uncovers some of , and uncovers some of , and uncovers some of those "unknown those "unknown those "unknown unknowns," where unknowns," where unknowns," where of course there are things of course there are things of course there are things you don't know that you don't know that you don't know that you don't know about, you don't know about, you don't know about, and things you and things you and things you know that you do know that you do know that you do n't know about. This is a very n't know about. This is a very n't know about. This is a very confusing sentence, confusing sentence, confusing sentence, but essentially, these are things but essentially, these are things but essentially, these are things you wouldn't you wouldn't you wouldn't guess but are guess but are guess but are worth taking care of, worth taking care of, worth taking care of, right? It could be right? It could be right? It could be security, security, security, deployment method, or deployment method, or deployment method, or just things just things just things you haven't thought about. And if you haven't thought about. And if you haven't thought about. And if you're not paying attention you're not paying attention you're not paying attention to them, the agent will just to them, the agent will just to them, the agent will just make something up or make something up or make something up or say, "Obviously, it's say, "Obviously, it's say, "Obviously, it's not important because they not important because they not important because they didn't mention it," and didn't mention it," and didn't mention it," and you'll end up with you'll end up with you'll end up with strange things strange things strange things or something completely or something completely or something completely different from what you different from what you different from what you imagined. imagined. Mhm. So the skill here is to just get it all out of just get it all out of your head. your head. Yes, the skill, I Yes, the skill, I believe, is believe, is believe, is knowing that you knowing that you ...Mhm. What is ...Mhm. What is ...Mhm. What is mastery?

mastery? mastery? Skill? Skill? Skill? Mastery is about Mastery is about Mastery is about being aware of the being aware of the being aware of the unknown unknowns, unknown unknowns, unknown unknowns, right? Understanding that right? Understanding that right? Understanding that you may you may you may miss something in your miss something in your miss something in your design and not be design and not be design and not be sure that you sure that you sure that you have created a perfect have created a perfect have created a perfect project. Um, and when you project. Um, and when you project. Um, and when you write code based on write code based on write code based on some idea, it's some idea, it's some idea, it's really hard, isn't really hard, isn't really hard, isn't it? It's a tricky thing it? It's a tricky thing it? It's a tricky thing because when you turn it because when you turn it because when you turn it into code, you notice the into code, you notice the into code, you notice the weaknesses in the weaknesses in the weaknesses in the design, and sometimes you design, and sometimes you design, and sometimes you can't see them can't see them can't see them until you get to until you get to until you get to writing the code, because writing the code, because writing the code, because it's kind of it's kind of it's kind of translating from translating from translating from fuzzy human fuzzy human fuzzy human thinking into thinking into thinking into executable executable executable instructions, right? There are always difficulties with instructions, right? There are always difficulties with instructions, right? There are always difficulties with things that need to be things that need to be things that need to be interconnected, and in interconnected, and in interconnected, and in this translation process this translation process . Certainly. I Certainly. I remember when I started remember when I started remember when I started creating creating creating automation. I automation. I automation. I tried to do my tried to do my tried to do my best to best to best to anticipate all the anticipate all the anticipate all the non-standard non-standard non-standard situations and, you know, situations and, you know, situations and, you know, create create create safeguards to safeguards to safeguards to protect against them. But protect against them. But protect against them. But then you put everything into then you put everything into then you put everything into operation, and all sorts of things happen operation, and all sorts of things happen operation, and all sorts of things happen , especially when the , especially when the , especially when the input data for the input data for the input data for the system is people, because system is people, because system is people, because people are just as people are just as people are just as unpredictable, unpredictable, unpredictable, if not more so, than if not more so, than if not more so, than agents. So I agents. So I agents. So I remember at least being remember at least being remember at least being able to communicate that to able to communicate that to able to communicate that to the client, saying, "Hey, the client, saying, "Hey, the client, saying, "Hey, I've built all these I've built all these I've built all these protections, but in the protections, but in the protections, but in the real world, real world, things can still happen, and so I'm going to be things can still happen, and so I'm going to be here to monitor the here to monitor the here to monitor the system and see what's system and see what's system and see what's happening, happening, happening, ensuring that all these ensuring that all these ensuring that all these failures happen in a failures happen in a failures happen in a safe manner."

safe manner." safe manner." Many in my Many in my Many in my audience strive audience strive audience strive to create solutions to create solutions to create solutions for their clients for their clients for their clients or stakeholders at or stakeholders at or stakeholders at work. And there's one thing work. And there's one thing work. And there's one thing I often notice— I often notice— it's a certain it's a certain it's a certain trust gap, because it's trust gap, because it's trust gap, because it's amazing how much you amazing how much you amazing how much you can build in can build in can build in one day and how one day and how one day and how expert you can expert you can expert you can seem if you seem if you seem if you lock yourself in a room lock yourself in a room lock yourself in a room for the weekend and for the weekend and for the weekend and come out on Monday with a come out on Monday with a come out on Monday with a finished portfolio. How do finished portfolio. How do finished portfolio. How do you think you should you think you should you think you should position yourself, position yourself, position yourself, your experience, your your experience, your your experience, your expertise, and your expertise, and your expertise, and your project portfolio? project portfolio? project portfolio? How do you position How do you position How do you position yourself in front of a yourself in front of a yourself in front of a potential client potential client to convince them to convince them to convince them that you are the right that you are the right that you are the right person person person for the for the for the job and to job and to job and to stand out from the 20 stand out from the 20 stand out from the 20 other vendors other vendors other vendors or or or promotion candidates they are promotion candidates they are promotion candidates they are considering? How do considering? How do considering? How do you stand out in a world you stand out in a world you stand out in a world where anyone can where anyone can where anyone can create things so create things so create things so quickly? quickly? quickly? Hurry up, friends, Hurry up, friends, Hurry up, friends, I have a completely I have a completely I have a completely free kit free kit free kit that I want to that I want to that I want to give to all of you give to all of you give to all of you to help you to help you to help you build your own AI OS. build your own AI OS. Whether you need to Whether you need to create it or create it or create it or scale it, this is the scale it, this is the scale it, this is the complete kit. It is complete kit. It is complete kit. It is based on my based on my based on my own structure, own structure, own structure, which I call the Herc Loop: which I call the Herc Loop: which I call the Herc Loop: context, connections, context, connections, context, connections, possibilities, and rhythm.

possibilities, and rhythm. possibilities, and rhythm. This thing has six This thing has six This thing has six skills. It skills. It skills. It helps you helps you helps you tune your intelligence tune your intelligence . She helps you . She helps you . She helps you prioritize prioritize prioritize your business and your business and your business and conducts the interview for you conducts the interview for you conducts the interview for you . This is exactly . This is exactly . This is exactly how I built how I built how I built my own my own AI-based operating system, and it's why AI-based operating system, and it's why I move so I move so I move so fast. So if you fast. So if you fast. So if you need it, need it, need it, the link to it is the link to it is the link to it is in the description. But, in in the description. But, in any case, any case, any case, let's get back to the video. let's get back to the video. let's get back to the video. Honestly, I've Honestly, I've Honestly, I've never been in this never been in this never been in this situation, you know? I have situation, you know? I have situation, you know? I have never been an never been an never been an independent independent independent consultant. Although consultant. Although consultant. Although no, I guess I was still an no, I guess I was still an no, I guess I was still an independent independent independent consultant. consultant. consultant. I just had a I just had a I just had a huge audience huge audience huge audience while I was doing it, while I was doing it, while I was doing it, you know? So I had a you know? So I had a you know? So I had a certain reputation certain reputation certain reputation that preceded me. that preceded me. that preceded me. I think people in I think people in I think people in these situations these situations these situations really want really want really want someone they can someone they can someone they can work with and someone who work with and someone who work with and someone who understands what's important to them understands what's important to them understands what's important to them , right? Someone , right? Someone , right? Someone who will try to bridge the who will try to bridge the who will try to bridge the communication communication communication gap. And since gap. And since gap. And since agents allow agents allow agents allow you to release you to release you to release prototypes, just... prototypes, just... Releasing—I would say " " vibe coding" at this point—is vibe coding" at this point—is vibe coding" at this point—is actually good, right? actually good, right? actually good, right? Because you're just Because you're just Because you're just trying to get to the trying to get to the trying to get to the end point end point end point pretty quickly. If pretty quickly. If pretty quickly. If you can create you can create you can create something that demonstrates to something that demonstrates to something that demonstrates to them that you understand them that you understand them that you understand their business, that you their business, that you their business, that you care about what's care about what's care about what's important to them, and that important to them, and that important to them, and that you can show you can show you can show real ingenuity real ingenuity real ingenuity in creating something in creating something in creating something real for them real for them real for them very quickly, I think very quickly, I think very quickly, I think they'll love it.

they'll love it. they'll love it. So I can't give So I can't give So I can't give specific advice specific advice specific advice on that because I've never on that because I've never on that because I've never been in such a been in such a been in such a competitive competitive competitive situation where you have to situation where you have to situation where you have to compete for roles and compete for roles and compete for roles and so on. But so on. But so on. But use AI to use AI to use AI to your advantage: if your advantage: if your advantage: if people are scared off by the fact that you people are scared off by the fact that you people are scared off by the fact that you can do can do can do the job in a day, the job in a day, the job in a day, just bill them just bill them just bill them for 5 days and for 5 days and for 5 days and rest for four rest for four rest for four of them. I think people of them. I think people of them. I think people don't yet realize don't yet realize don't yet realize how quickly you can how quickly you can how quickly you can create create create high-quality things high-quality things high-quality things with agents, with agents, with agents, especially those who aren't " especially those who aren't " hooked" on AI. So hooked" on AI. So hooked" on AI. So I think you just have to I think you just have to I think you just have to adapt what adapt what adapt what you do for them. you do for them. What's interesting about what you What's interesting about what you said is that it's said is that it's said is that it's humanity that humanity that humanity that helps build helps build helps build trust, not " trust, not " look at how many look at how many look at how many cool things I've cool things I've cool things I've created" or " created" or " look at how many complex look at how many complex look at how many complex structures I know." It's a matter of structures I know." It's a matter of structures I know." It's a matter of humanity: humanity: humanity: does this person care about what I does this person care about what I does this person care about what I care about, and does care about, and does care about, and does he think about he think about he think about protecting my business, protecting my business, protecting my business, not just about not just about not just about getting a paycheck from me? getting a paycheck from me? getting a paycheck from me? What about What about What about your students? I'm your students? I'm your students? I'm sure you've seen sure you've seen sure you've seen people about whom you can people about whom you can people about whom you can say: they have say: they have say: they have potential, they are very potential, they are very potential, they are very talented. And you talented. And you talented. And you might have noticed traits might have noticed traits might have noticed traits that make it that make it that make it clear: this person clear: this person clear: this person needs more time, a needs more time, a needs more time, a few more weeks of few more weeks of few more weeks of work. What exactly work. What exactly work. What exactly catches your eye in catches your eye in catches your eye in your students?

your students? I think the truly I think the truly successful people are the ones successful people are the ones successful people are the ones who are the best at who are the best at who are the best at communicating, communicating, communicating, right? This has always been right? This has always been right? This has always been true for both true for both true for both developers and developers and developers and non-developers. non-developers. Everyone I've seen at the Everyone I've seen at the highest level of highest level of highest level of development has always been development has always been development has always been incredibly adept incredibly adept incredibly adept at communicating. This at communicating. This at communicating. This means that when they means that when they means that when they make changes to the code, make changes to the code, make changes to the code, they write a very they write a very they write a very detailed report about detailed report about exactly what changed. exactly what changed. exactly what changed. It's easy to read, It's easy to read, It's easy to read, easy to understand, and easy to understand, and easy to understand, and their thinking is their thinking is their thinking is expressed very clearly. In expressed very clearly. In expressed very clearly. In other words, other words, other words, the clarity of their the clarity of their the clarity of their thinking thinking thinking is reflected in is reflected in is reflected in the product they the product they the product they create. And what I create. And what I create. And what I notice in those who notice in those who notice in those who don't do this is that don't do this is that they get worse they get worse they get worse results because they can't results because they can't results because they can't clearly clearly clearly set the task for the set the task for the set the task for the agent and explain what agent and explain what agent and explain what exactly is going wrong. They exactly is going wrong. They exactly is going wrong. They do not think in clear do not think in clear do not think in clear categories and do not categories and do not categories and do not have clear boundaries. So have clear boundaries. So have clear boundaries. So actually I think actually I think actually I think the quality of your the quality of your the quality of your thinking and the quality of thinking and the quality of thinking and the quality of your language are your language are your language are key, right? key, right? key, right? Not only when you Not only when you Not only when you communicate with communicate with communicate with the customer, but also when you the customer, but also when you the customer, but also when you explain to the agent explain to the agent explain to the agent exactly what the customer wants exactly what the customer wants . I have some . I have some . I have some skills for this, and skills for this, and an idea from programming is also woven into this— the concept of DDD, or the concept of DDD, or Object- Object- Oriented Oriented Oriented Design.

Design. Design. It sounds very difficult. It sounds very difficult. But it is based on But it is based on the idea that the idea that the idea that when creating when creating when creating a project, you a project, you a project, you use the so- use the so- use the so- called "universal called "universal called "universal language": the terms in the code language": the terms in the code language": the terms in the code should match should match should match those used by those used by those used by developers to describe the developers to describe the developers to describe the product, by the client product, by the client product, by the client asking to create this asking to create this asking to create this product, and by the product, and by the product, and by the users working users working users working with this product. with this product. So, if you have the So, if you have the concept of "order" concept of "order" , right? What is , right? What is , right? What is an order? It should an order? It should an order? It should look the same look the same look the same in the code, in the database, in the code, in the database, in the code, in the database, in the interface, and in the in the interface, and in the in the interface, and in the conversation between you and conversation between you and conversation between you and the client. And actually, the client. And actually, the client. And actually, iterating on this, iterating on this, iterating on this, building this, and the building this, and the building this, and the precision of this language precision of this language precision of this language again again again helps you helps you helps you align what align what align what they actually they actually they actually want with what you're want with what you're want with what you're actually building. I actually building. I actually building. I can't recommend can't recommend can't recommend this technique enough. By the way, this technique enough. By the way, this technique enough. By the way, I have a modified I have a modified I have a modified version of “Grill Me” that version of “Grill Me” that version of “Grill Me” that does this, so it does this, so it does this, so it helps you helps you helps you come up with these come up with these come up with these terms, it’s terms, it’s terms, it’s called “Grill with called “Grill with called “Grill with Docs.” That's what Docs.” That's what Grill with Docs does. This Grill with Docs does. This Grill with Docs does. This helps you helps you helps you create this create this create this subject language as you subject language as you subject language as you work. I work. I work. I mean, I don't know. I'm a mean, I don't know. I'm a mean, I don't know. I'm a little bit, I'm little bit, I'm little bit, I'm pushing DDD a little bit, but it's a pushing DDD a little bit, but it's a pushing DDD a little bit, but it's a pretty advanced pretty advanced pretty advanced approach. You know, you approach. You know, you approach. You know, you talk to a lot of talk to a lot of talk to a lot of programmers, they have programmers, they have programmers, they have n't heard of DDD, but I n't heard of DDD, but I n't heard of DDD, but I think it's think it's think it's extraordinary for extraordinary for extraordinary for developing with agents.

developing with agents. developing with agents. Interesting. What, what do you Interesting. What, what do you Interesting. What, what do you think about JDD? Have you ever think about JDD? Have you ever think about JDD? Have you ever heard of heard of heard of such a thing? such a thing? such a thing? No, what is JDD? No, what is JDD? Jealousy-driven Jealousy-driven development. Say to Claude: " development. Say to Claude: " Hey, hey, Astro does it Hey, hey, Astro does it Hey, hey, Astro does it much better than much better than much better than you." What is your you." What is your you." What is your problem? What is problem? What is problem? What is going on here? going on here? going on here? There's spectral There's spectral There's spectral development, there's development, there's development, there's readme-driven development— readme-driven development— it's fun. Uh, it's fun. Uh, behavior-driven development (BDD), I behavior-driven development (BDD), I mean, it's everywhere. mean, it's everywhere. mean, it's everywhere. That, and I just saw That, and I just saw That, and I just saw that Anthropic announced that Anthropic announced that Anthropic announced some kind of consequences, some kind of consequences, some kind of consequences, or that it would prohibit or that it would prohibit or that it would prohibit you from being cruel you from being cruel you from being cruel or mean to Claude. I don't or mean to Claude. I don't or mean to Claude. I don't know if you saw it. know if you saw it. know if you saw it. I just saw it last night I just saw it last night I just saw it last night , so , so , so we'll see what we'll see what we'll see what comes of it. But you've comes of it. But you've comes of it. But you've talked a lot on YouTube, here, talked a lot on YouTube, here, talked a lot on YouTube, here, on stages about your on stages about your on stages about your skills. And there's a reason skills. And there's a reason skills. And there's a reason why I why I why I use them, there's a use them, there's a use them, there's a reason why so reason why so reason why so many people many people many people use them. And what use them. And what use them. And what I'd like to hear from I'd like to hear from I'd like to hear from you, before all of this, you, before all of this, you, before all of this, you know, I'm going to you know, I'm going to you know, I'm going to put you in a little bit of an put you in a little bit of an put you in a little bit of an awkward position. I would awkward position. I would awkward position. I would really like to hear really like to hear really like to hear some story from the " some story from the " some story from the " battlefield." You know, I battlefield." You know, I battlefield." You know, I really enjoy really enjoy really enjoy learning from people, learning from people, learning from people, hearing about the failures hearing about the failures hearing about the failures they've had or they've had or they've had or why they learned a why they learned a why they learned a lesson. And I'm sure there lesson. And I'm sure there lesson. And I'm sure there was a point was a point was a point where you decided, " where you decided, " Okay, I need to Okay, I need to Okay, I need to develop a skill develop a skill develop a skill around this so I around this so I around this so I don't have to face don't have to face don't have to face this problem again." Or this problem again." Or this problem again." Or you had some, you had some, you had some, you know, big you know, big you know, big problem in your problem in your problem in your projects or with projects or with projects or with students. I would really students. I would really students. I would really like to peek like to peek like to peek into your brain a little and into your brain a little and into your brain a little and understand one of these understand one of these understand one of these stories from which you stories from which you stories from which you learned an important learned an important learned an important lesson.

lesson. Well, I was really Well, I was really interested in a talk interested in a talk interested in a talk I saw from a I saw from a I saw from a guy from OpenAI. Uh, it guy from OpenAI. Uh, it guy from OpenAI. Uh, it was on AI Engineer, I, uh, was on AI Engineer, I, uh, was on AI Engineer, I, uh, Sean Grove, I think Sean Grove, I think Sean Grove, I think he said that he said that he said that the specs—that's the the specs—that's the the specs—that's the new code, right? new code, right? new code, right? Specifications are the Specifications are the Specifications are the new code. And the theory new code. And the theory new code. And the theory was that was that was that you don't you don't you don't have to write code anymore, have to write code anymore, have to write code anymore, because you can because you can because you can write a write a write a specification—just specification—just specification—just create a very create a very create a very detailed document, and detailed document, and detailed document, and the agents will help the agents will help the agents will help you with what you you with what you you with what you want that code to look like want that code to look like . And then you just . And then you just . And then you just give the agent the give the agent the give the agent the task of turning task of turning task of turning that specification into that specification into that specification into code. Right? And it's still a code. Right? And it's still a code. Right? And it's still a very, very very, very very, very popular idea. People popular idea. People popular idea. People love this thing, and there are love this thing, and there are tons of frameworks built around it. I would tons of frameworks built around it. I would say that this is say that this is say that this is essentially essentially essentially specification-driven development. specification-driven development. specification-driven development. It's a pretty It's a pretty It's a pretty vague term vague term , but at least that's how I , but at least that's how I , but at least that's how I define it define it define it . This is when you . This is when you . This is when you keep keep keep the specifications and the specifications and the specifications and just refine just refine just refine them as you go. And I them as you go. And I them as you go. And I tried to do it. tried to do it. tried to do it. I really tried. I really tried. I really tried. This was before This was before This was before I came up with my I came up with my I came up with my skills and stuff.

skills and stuff. skills and stuff. You know, I have a You know, I have a You know, I have a repository that is repository that is repository that is my pride, it's my my pride, it's my my pride, it's my own own own video editor. This is video editor. This is video editor. This is something like a version of something like a version of something like a version of Dscript. There's AI for Dscript. There's AI for Dscript. There's AI for writing text, writing text, writing text, thumbnail design, well, thumbnail design, well, thumbnail design, well, you know, I do you know, I do you know, I do absolutely everything on it. absolutely everything on it. And I tried to And I tried to turn it into a turn it into a turn it into a big spec, big spec, big spec, and tried to... and tried to... apply apply spec-driven development, I guess. spec-driven development, I guess. And what I found was: first of all And what I found was: first of all , I kept , I kept , I kept wanting to review the wanting to review the wanting to review the code because I was code because I was code because I was curious, and I could see the curious, and I could see the curious, and I could see the code getting code getting code getting worse and worse. And worse and worse. And worse and worse. And the agent produced the agent produced the agent produced worse and worse worse and worse worse and worse results. And I results. And I results. And I tried this for months tried this for months tried this for months because I thought: this guy because I thought: this guy because I thought: this guy works at OpenAI, right? works at OpenAI, right? works at OpenAI, right? He knows what he's He knows what he's He knows what he's talking about. I will be able to talking about. I will be able to talking about. I will be able to master this master this master this concept. And I just concept. And I just concept. And I just thought about it again, and thought about it again, and thought about it again, and again, and again. But it did again, and again. But it did again, and again. But it did n't work. So I n't work. So I n't work. So I realized: no, I'm realized: no, I'm realized: no, I'm going to abandon going to abandon going to abandon these specifications. these specifications. these specifications. I'm actually going to I'm actually going to I'm actually going to focus more focus more focus more on the code itself. And what on the code itself. And what on the code itself. And what pushed me to do pushed me to do pushed me to do this is that this is that this is that code is the environment in code is the environment in code is the environment in which the agent operates. which the agent operates. which the agent operates. Truth? As I said Truth? As I said Truth? As I said before. If you have an before. If you have an before. If you have an agent that essentially agent that essentially agent that essentially resembles that dog resembles that dog resembles that dog sitting in a sitting in a sitting in a burning house, it's not going burning house, it's not going burning house, it's not going to produce good code, to produce good code, to produce good code, right? The dog is right? The dog is right? The dog is in a bad in a bad in a bad environment, environment, environment, you know. I wanted you know. I wanted you know. I wanted good code, good code, good code, code that made sense, and code that made sense, and code that made sense, and if you think about if you think about if you think about what " what " good code" means, it's a good code" means, it's a good code" means, it's a codebase that's codebase that's codebase that's easy to make changes to.

easy to make changes to. The change shouldn't The change shouldn't mess up something the mess up something the mess up something the agent didn't think of. A agent didn't think of. A code base where everything is code base where everything is easy to modify. And that's what easy to modify. And that's what easy to modify. And that's what I started to I started to I started to focus on. And focus on. And focus on. And when I started when I started when I started focusing on focusing on focusing on that and going back to the that and going back to the that and going back to the code, that's when I felt like, code, that's when I felt like, code, that's when I felt like, okay, I'm finally making okay, I'm finally making okay, I'm finally making progress and I can get progress and I can get progress and I can get things moving things moving things moving . But the . But the . But the scope of development scope of development scope of development just didn't suit me. just didn't suit me. The models are getting The models are getting better every better every better every time. Every time a time. Every time a time. Every time a new iteration comes out new iteration comes out new iteration comes out , they , they , they get better. Why do you get better. Why do you get better. Why do you think these think these think these models still don't models still don't models still don't understand many of understand many of understand many of these concepts or these concepts or these concepts or implement them implement them implement them natively? And when do you natively? And when do you natively? And when do you think that think that think that will happen? will happen? So it's a So it's a communication communication communication barrier, right? barrier, right? barrier, right? Models can't Models can't Models can't know what's bothering you know what's bothering you know what's bothering you until you until you until you tell them. And skills tell them. And skills are one way are one way are one way to let models know to let models know to let models know what's important to you what's important to you what's important to you . Now I . Now I . Now I think that the models think that the models think that the models are getting better, and are getting better, and are getting better, and the environments for their the environments for their the environments for their use (harnesses) use (harnesses) use (harnesses) are also improving.

are also improving. are also improving. I think people I think people I think people underestimate the impact of underestimate the impact of underestimate the impact of tools like this tools like this tools like this on models, you on models, you on models, you know, Claude Code, Claude.ai. know, Claude Code, Claude.ai. know, Claude Code, Claude.ai. And Codex, or something like that. And Codex, or something like that. And Codex, or something like that. I think models I think models I think models will get better, of course, will get better, of course, will get better, of course, but they'll but they'll but they'll never be never be never be good enough good enough good enough to read your mind to read your mind , right? They , right? They , right? They will never be perfectly will never be perfectly will never be perfectly in sync with in sync with in sync with you every time. So you you every time. So you you every time. So you need some need some need some structure, some structure, some structure, some process, to help process, to help process, to help that brilliant that brilliant that brilliant developer understand developer understand what you want, because what you want, because what you want, because building a building a building a great product great product great product requires both a good requires both a good requires both a good manager and a good manager and a good manager and a good developer. You should developer. You should developer. You should take care of the matter. take care of the matter. take care of the matter. You have to look You have to look You have to look at the results and at the results and at the results and add your add your add your own flavor to them. So I own flavor to them. So I own flavor to them. So I always believe that there has to always believe that there has to always believe that there has to be some level of be some level of be some level of process on top of that, process on top of that, process on top of that, just to overcome just to overcome just to overcome that communication that communication that communication barrier. This is logical. barrier. This is logical. barrier. This is logical. That makes sense. That makes sense. That makes sense. How much faster do How much faster do How much faster do you think you can you think you can you think you can move now move now move now after learning these basics, after learning these basics, after learning these basics, compared to compared to compared to when you were doing everything when you were doing everything when you were doing everything manually? Listen, that's manually? Listen, that's manually? Listen, that's just an incredible just an incredible just an incredible difference, right? It's difference, right? It's difference, right? It's just an incredible just an incredible just an incredible amount of stuff amount of stuff amount of stuff I can release. It's I can release. It's I can release. It's simply impossible simply impossible simply impossible to compare. I to compare. I to compare. I currently have about four currently have about four currently have about four agents working agents working agents working on different on different on different projects. I gave them projects. I gave them projects. I gave them all long-term all long-term all long-term tasks. I said, " tasks. I said, " Okay, do this, this, Okay, do this, this, Okay, do this, this, and this." Here are my and this." Here are my and this." Here are my long-term goals.

long-term goals. long-term goals. Go and deal with Go and deal with Go and deal with it." I have it." I have it." I have new skills that new skills that new skills that I'm experimenting with, I'm experimenting with, I'm experimenting with, like something like something like something like a " like a " chief of staff" skill that chief of staff" skill that chief of staff" skill that manages sub-agents, and, manages sub-agents, and, manages sub-agents, and, you know, it's just you know, it's just you know, it's just incredible. It's like a incredible. It's like a incredible. It's like a team of eight team of eight team of eight developers that developers that developers that I have access to, and they I have access to, and they I have access to, and they work for me, work for me, work for me, you know? I you know? I you know? I check them out, check them out, check them out, make sure their make sure their make sure their plans are good, plans are good, plans are good, arrange for them to be "interrogated arrange for them to be "interrogated " and all that. And I...I don't " and all that. And I...I don't " and all that. And I...I don't know. I mean, you can know. I mean, you can know. I mean, you can look at my GitHub look at my GitHub , the number of , the number of , the number of commits I commits I commits I push out, it's, you know push out, it's, you know , 200, 300 a day or something like that , 200, 300 a day or something like that , 200, 300 a day or something like that . Whereas before, . Whereas before, . Whereas before, you know, 20 a day would have been a you know, 20 a day would have been a you know, 20 a day would have been a good good good result. But result. But result. But again, you again, you again, you can can can check it out for yourself. So, it's check it out for yourself. So, it's check it out for yourself. So, it's not comparable. Even not comparable. Even not comparable. Even remotely. remotely. remotely. I have to ask about your I have to ask about your I have to ask about your favorite tool favorite tool favorite tool and model. and model. I am currently using I am currently using Claude Code and Opus 5.5. Opus 5.5 is a Claude Code and Opus 5.5. Opus 5.5 is a great model. I great model. I great model. I use it at a use it at a use it at a much lower much lower much lower effort level than effort level than effort level than most people, and this is most people, and this is most people, and this is one of my one of my one of my provocative thoughts: provocative thoughts: provocative thoughts: I think I think I think most people most people most people set their effort set their effort set their effort level too high level too high level too high . I think . I think people look at the people look at the people look at the effort and think, " effort and think, " Yeah, I'm going to Yeah, I'm going to Yeah, I'm going to max it out, right?"

max it out, right?" max it out, right?" Like, I want Like, I want Like, I want him to make an effort, him to make an effort, him to make an effort, you know? And I you know? And I you know? And I use Opus 5.5 at a use Opus 5.5 at a use Opus 5.5 at a medium medium medium effort level, actually. And I'm effort level, actually. And I'm effort level, actually. And I'm even thinking about going even thinking about going even thinking about going to a low level, because to a low level, because to a low level, because it gives much it gives much it gives much greater greater greater token efficiency. It token efficiency. It token efficiency. It responds much responds much responds much faster, and because I've faster, and because I've faster, and because I've raised the bar for raised the bar for raised the bar for my agents, they don't my agents, they don't my agents, they don't have to work have to work have to work as hard to as hard to as hard to write good write good write good code—they just code—they just code—they just write what write what write what passes the tests, and the passes the tests, and the passes the tests, and the tests tests tests check everything. So that's what check everything. So that's what check everything. So that's what I I I like right now, like right now, like right now, but I'm not that much of a but I'm not that much of a but I'm not that much of a polyglot. I don't polyglot. I don't polyglot. I don't test all these test all these test all these models and models and models and tools because I tools because I tools because I believe that while the believe that while the believe that while the model and environment are model and environment are model and environment are important for important for important for quality work, I quality work, I quality work, I also believe: also believe: also believe: just improve just improve just improve the conditions in which the conditions in which the conditions in which your agents work. your agents work. That's where I find the That's where I find the most benefit. most benefit. I recently moved a I recently moved a project that I did project that I did project that I did n't know how to do before n't know how to do before n't know how to do before so that an agent could go into a so that an agent could go into a so that an agent could go into a browser and browser and browser and view it. It view it. It view it. It was a complex project.

was a complex project. was a complex project. It depended on It depended on It depended on many third-party many third-party many third-party libraries, for example, libraries, for example, libraries, for example, it required it required it required OBS to be run locally OBS to be run locally . This was my . This was my . This was my video editor. How do video editor. How do video editor. How do I get it to I get it to I get it to work so it work so it work so it can do this while I'm can do this while I'm can do this while I'm recording video at the same time recording video at the same time recording video at the same time ? It was quite ? It was quite ? It was quite difficult. But I difficult. But I difficult. But I fixed it, and now the fixed it, and now the fixed it, and now the agent can agent can agent can follow follow follow links, links, links, take screenshots take screenshots , actually , actually , actually run run run the program properly, and that the program properly, and that the program properly, and that change alone has led to a change alone has led to a change alone has led to a huge huge huge improvement in the quality of improvement in the quality of improvement in the quality of what I can what I can what I can create because the create because the create because the agent has much agent has much agent has much better access to the better access to the better access to the data. And I think I could data. And I think I could data. And I think I could get, get, get, you know, just as you know, just as you know, just as good a result with good a result with good a result with Opus 4.5 and that kind of Opus 4.5 and that kind of Opus 4.5 and that kind of setup, setup, setup, you know what I mean? you know what I mean? you know what I mean? It's just an incredible It's just an incredible It's just an incredible difference it made. difference it made. difference it made. And so, honestly, And so, honestly, And so, honestly, I think working on the I think working on the I think working on the environment that environment that environment that your agents work in your agents work in pays off very quickly. And I know that's kind of an evasion evasion , right? Because you , right? Because you , right? Because you want me to give a want me to give a want me to give a heated comment, heated comment, heated comment, saying Codex saying Codex saying Codex sucks and local sucks and local sucks and local code is great, but code is great, but code is great, but I actually don't care I actually don't care I actually don't care about the about the about the testing system itself. It testing system itself. It testing system itself. It really doesn't matter, really doesn't matter, really doesn't matter, you know, she's you know, she's you know, she's improving, and that's improving, and that's improving, and that's great. And the Opus 5.5, great. And the Opus 5.5, great. And the Opus 5.5, in particular, is a much in particular, is a much in particular, is a much better model than better model than better model than the fifth, much more the fifth, much more the fifth, much more efficient. But efficient. But efficient. But in reality, I treat in reality, I treat in reality, I treat them as them as them as employees of my employees of my employees of my company, and I want company, and I want company, and I want my company to be my company to be my company to be the best it can be, so that the the best it can be, so that the the best it can be, so that the basic level of basic level of basic level of employees is employees is employees is higher.

higher. higher. Yeah, you know, I Yeah, you know, I Yeah, you know, I actually really actually really actually really like that like that like that answer because I think answer because I think answer because I think it helps people it helps people it helps people realize: you don't realize: you don't realize: you don't have to have to have to worry so much about worry so much about worry so much about the newest model and the newest model and the newest model and making sure you making sure you making sure you fully understand what fully understand what fully understand what it's best at and it's best at and it's best at and all these benchmarks. Because, all these benchmarks. Because, all these benchmarks. Because, you know, if you think back to you know, if you think back to you know, if you think back to the days before the Opus, when the days before the Opus, when the days before the Opus, when we had the Sonnet, it was a we had the Sonnet, it was a we had the Sonnet, it was a very, very good very, very good very, very good model. model. model. A lot of things could be automated A lot of things could be automated A lot of things could be automated . You could . You could . You could perform a lot of perform a lot of perform a lot of intellectual intellectual intellectual tasks and do a tasks and do a tasks and do a lot of things with this lot of things with this lot of things with this model. So even the model. So even the model. So even the Opus 5.5 on low Opus 5.5 on low Opus 5.5 on low settings is better settings is better settings is better than that model, and I than that model, and I than that model, and I think a lot of people think a lot of people think a lot of people don't realize that don't realize that don't realize that because of how quickly because of how quickly because of how quickly things change. And you things change. And you things change. And you know, I've always know, I've always know, I've always believed that someone believed that someone believed that someone like you, Matt, with a like you, Matt, with a like you, Matt, with a very old model would very old model would very old model would still be able to produce a still be able to produce a still be able to produce a better result than better result than better result than someone with Opus 5.5 who someone with Opus 5.5 who someone with Opus 5.5 who has no idea how to has no idea how to has no idea how to communicate. So that's what communicate. So that's what communicate. So that's what I think is I think is I think is really cool about this, and really cool about this, and really cool about this, and I really I really I really like that like that like that answer. And now the other thing answer. And now the other thing answer. And now the other thing I'm I'm I'm interested in about you is, interested in about you is, someone might have someone might have listened to this whole listened to this whole listened to this whole podcast and still podcast and still podcast and still feel feel feel overwhelmed by overwhelmed by overwhelmed by a lot of the things a lot of the things a lot of the things we're talking about, but you we're talking about, but you we're talking about, but you know, there are a lot of know, there are a lot of know, there are a lot of benefits, and you were benefits, and you were benefits, and you were talking about talking about talking about how much faster you've gotten how much faster you've gotten how much faster you've gotten , and we were , and we were , and we were talking about talking about talking about how much lower the how much lower the how much lower the barrier to entry has become.

barrier to entry has become. Whatever your Whatever your final motivational final motivational final motivational message, your message, your message, your final push to final push to final push to encourage someone to take encourage someone to take encourage someone to take action, get down action, get down action, get down to work, get their to work, get their to work, get their hands dirty, and learn hands dirty, and learn hands dirty, and learn more. more. So I think if So I think if you have a desire to get you have a desire to get you have a desire to get better at this, I have to better at this, I have to better at this, I have to say that my say that my say that my skills are a great skills are a great skills are a great place to start. place to start. place to start. You know, if you're You know, if you're You know, if you're looking for, well, just looking for, well, just looking for, well, just give me a basic give me a basic give me a basic idea of ​​what idea of ​​what idea of ​​what to do, then to do, then to do, then downloading my downloading my downloading my skills or any skills or any skills or any tool—just tool—just tool—just pick one of the pick one of the pick one of the main ones. Choose Claude main ones. Choose Claude main ones. Choose Claude Code, Codex, GitHub Copilot, or Code, Codex, GitHub Copilot, or Code, Codex, GitHub Copilot, or whatever. You know, don't whatever. You know, don't whatever. You know, don't spend too spend too spend too much time much time much time thinking about the thinking about the thinking about the instrument. Just instrument. Just instrument. Just try it. And then, try it. And then, try it. And then, after that, it's after that, it's after that, it's about...I recently about...I recently about...I recently interviewed Lauren Tan, the interviewed Lauren Tan, the interviewed Lauren Tan, the Potato on Twitter, who works Potato on Twitter, who works Potato on Twitter, who works at GrokBot, and she at GrokBot, and she at GrokBot, and she talks about climbing the talks about climbing the talks about climbing the trust ladder, right? trust ladder, right? trust ladder, right? Climbing the Climbing the Climbing the trust ladder because you won't trust ladder because you won't trust ladder because you won't trust your trust your trust your agents when agents when agents when you see them you see them you see them working for the first time. And that's working for the first time. And that's working for the first time. And that's good. This is the right good. This is the right good. This is the right impulse. You impulse. You impulse. You shouldn't trust shouldn't trust shouldn't trust this blindly. You this blindly. You this blindly. You have to, well, you know, have to, well, you know, have to, well, you know, when you learn when you learn when you learn to drive, which pedal do you to drive, which pedal do you to drive, which pedal do you learn first? Are you learn first? Are you learn first? Are you learning learning learning to use to use to use the accelerator or the the accelerator or the the accelerator or the brake? You're learning brake? You're learning brake? You're learning to use to use to use the brake, right? You the brake, right? You the brake, right? You learn to stop learn to stop learn to stop because you don't want to because you don't want to because you don't want to just push and just push and just push and push because you'll push because you'll push because you'll push past your push past your push past your limits, create things you did limits, create things you did n't want to see n't want to see , and not feel what you're , and not feel what you're , and not feel what you're using. So using. So using. So I recommend just I recommend just I recommend just trying to open trying to open trying to open the model, give it the model, give it the model, give it a few queries, a few queries, a few queries, try opening an try opening an try opening an empty folder, empty folder, empty folder, use VS Code use VS Code use VS Code as well, because sometimes it's as well, because sometimes it's as well, because sometimes it's nice to be able to

nice to be able to nice to be able to peek peek peek inside the code. And inside the code. And from there you can from there you can start working on start working on start working on some idea. It really some idea. It really some idea. It really helps to have helps to have helps to have an idea of ​​what you an idea of ​​what you an idea of ​​what you want to create and to want to create and to want to create and to have something interesting in your head that you have something interesting in your head that you have something interesting in your head that you can actually work towards can actually work towards can actually work towards . Because . Because . Because if you don't have a if you don't have a if you don't have a reason or mission for reason or mission for reason or mission for doing it, it's much doing it, it's much doing it, it's much harder to find harder to find harder to find motivation and your feedback loop motivation and your feedback loop won't be as strong . Because when you learn and . Because when you learn and . Because when you learn and see a cool see a cool see a cool idea come to idea come to idea come to life, it's simple, it's life, it's simple, it's life, it's simple, it's like a multiplier. So from like a multiplier. So from like a multiplier. So from that point on, you that point on, you that point on, you start to watch start to watch start to watch the agent like a hawk, the agent like a hawk, the agent like a hawk, understand everything they understand everything they understand everything they do, understand everything the do, understand everything the toolkit gives you, and you toolkit gives you, and you start to feel start to feel start to feel when you trust the when you trust the when you trust the agent and when you agent and when you agent and when you understand agents. And understand agents. And understand agents. And when you get to when you get to when you get to that level, you that level, you that level, you can start can start can start using using using maybe two agents, maybe two agents, maybe two agents, or you can start or you can start or you can start writing some writing some writing some skills to make the skills to make the skills to make the agent do agent do agent do non-standard things. You non-standard things. You non-standard things. You can start can start can start connecting it to connecting it to connecting it to different data sources.

different data sources. different data sources. You can connect You can connect You can connect it to your Slack, to it to your Slack, to it to your Slack, to Twitter, well, to Twitter, well, to Twitter, well, to whatever, to your Gmail whatever, to your Gmail . And you can start with this . And you can start with this . And you can start with this , you know, I , you know, I , you know, I have an agent who manages my to- have an agent who manages my to- have an agent who manages my to- do list for do list for do list for me. So I have, me. So I have, me. So I have, you know, I dictate you know, I dictate you know, I dictate something into my watch something into my watch , it signals the , it signals the , it signals the system, it goes system, it goes system, it goes into my to-do list, and into my to-do list, and into my to-do list, and then the agent sorts it then the agent sorts it then the agent sorts it and stuff. So you and stuff. So you and stuff. So you start start start integrating it into your integrating it into your integrating it into your life and into your life and into your life and into your work. That's all. work. That's all. work. That's all. Climb the ladder of Climb the ladder of Climb the ladder of trust, start trust, start trust, start small, start with small, start with small, start with my skills, if my skills, if my skills, if you want, or not, you know you want, or not, you know . And from that point on, you . And from that point on, you . And from that point on, you can just can just can just go crazy with it, and it's go crazy with it, and it's go crazy with it, and it's amazing how, if you amazing how, if you amazing how, if you just pay just pay just pay attention to it and attention to it and attention to it and try to think try to think try to think on its level, it's amazing on its level, it's amazing on its level, it's amazing how quickly you can how quickly you can how quickly you can climb that climb that climb that ladder of trust. ladder of trust. ladder of trust. I like this I like this I like this answer. I kind of answer. I kind of answer. I kind of call it this in a lot of call it this in a lot of call it this in a lot of my videos, or this my videos, or this my videos, or this theory. I sometimes theory. I sometimes theory. I sometimes call it teaching call it teaching call it teaching a child to ride a a child to ride a a child to ride a bike, the bike, the bike, the bicycle method, about how bicycle method, about how bicycle method, about how you have to earn you have to earn you have to earn that trust, and you don't that trust, and you don't that trust, and you don't start by start by start by trying to teach trying to teach trying to teach 10 kids at once. You 10 kids at once. You 10 kids at once. You have to keep a close eye have to keep a close eye have to keep a close eye on them, you on them, you on them, you correct them, and at correct them, and at correct them, and at the end of the day, you don't the end of the day, you don't the end of the day, you don't just let 20 just let 20 just let 20 kids ride kids ride kids ride bikes down the highway, you bikes down the highway, you bikes down the highway, you go go go inside to take a nap yourself.

inside to take a nap yourself. inside to take a nap yourself. You are still very much You are still very much You are still very much involved in this process involved in this process . So I . So I . So I like that like that like that answer, and you know, I've answer, and you know, I've answer, and you know, I've seen Elon, Boris, seen Elon, Boris, seen Elon, Boris, and Sam, I think, and and Sam, I think, and and Sam, I think, and probably a lot of other probably a lot of other probably a lot of other people, essentially people, essentially people, essentially say some say some say some version of that. Give version of that. Give version of that. Give Model X your Model X your Model X your most ambitious most ambitious most ambitious tasks or your tasks or your tasks or your most ambitious goals. And most ambitious goals. And most ambitious goals. And I'm really curious that I'm really curious that I'm really curious that they're coming out and they're coming out and they're coming out and saying this, because I saying this, because I , well, I totally , well, I totally , well, I totally believe it, but I'd like to believe it, but I'd like to believe it, but I'd like to hear from you. Why do you hear from you. Why do you hear from you. Why do you think they think they think they felt the need to felt the need to felt the need to tell the world, "Hey tell the world, "Hey , try our , try our , try our model, but model, but model, but test it on test it on test it on your most ambitious your most ambitious your most ambitious tasks"? I tasks"? I don't know, because I think the don't know, because I think the opposite, right? I opposite, right? I opposite, right? I think give her your think give her your think give her your most mundane, most mundane, most mundane, boring job, you know, boring job, you know, boring job, you know, give her the things you give her the things you give her the things you hate to do. Um hate to do. Um , , , for example, task her with for example, task her with for example, task her with preparing draft preparing draft preparing draft responses to people who responses to people who responses to people who ask you for something, ask you for something, ask you for something, and give her simple, and give her simple, and give her simple, achievable tasks where you achievable tasks where you achievable tasks where you can evaluate the can evaluate the can evaluate the result and result and result and how how how good it is. Of course, it good it is. Of course, it good it is. Of course, it can be can be can be used for used for used for brainstorming. It brainstorming. It brainstorming. It can be can be can be used to used to used to brainstorm ideas and brainstorm ideas and how to turn them into how to turn them into how to turn them into execution, and execution, and execution, and it's quite good at that.

it's quite good at that. it's quite good at that. In fact, it's helpful In fact, it's helpful In fact, it's helpful to use to use to use "Grill Me" for this. But I would "Grill Me" for this. But I would "Grill Me" for this. But I would say that the real say that the real say that the real benefit I benefit I benefit I get from AI, and the get from AI, and the get from AI, and the reason I reason I reason I use it, is that use it, is that use it, is that it it it helps helps helps automate boring automate boring automate boring things. This helps things. This helps things. This helps me stay focused on the me stay focused on the me stay focused on the strategic level strategic level strategic level and on what I and on what I and on what I want to focus on. want to focus on. You know, I create You know, I create video content and video content and video content and written content, and written content, and written content, and writing that content writing that content writing that content has become a lot easier has become a lot easier has become a lot easier with the advent of AI, because I with the advent of AI, because I with the advent of AI, because I load everything into Sonic load everything into Sonic load everything into Sonic 5.5, and it creates 5.5, and it creates 5.5, and it creates a draft. I'm a draft. I'm a draft. I'm finishing it, and it's been finishing it, and it's been finishing it, and it's been just great just great just great for me because the work has for me because the work has for me because the work has become more interesting, because become more interesting, because become more interesting, because before I didn't like before I didn't like before I didn't like writing this kind of stuff. It writing this kind of stuff. It writing this kind of stuff. It was real was real was real torture. I even torture. I even torture. I even paid people to paid people to paid people to do it for do it for do it for me. So yeah, me. So yeah, me. So yeah, give her your give her your give her your boring job. boring job. boring job. Assign her small, Assign her small, Assign her small, everyday tasks everyday tasks everyday tasks that you don't have to that you don't have to that you don't have to worry about if worry about if worry about if she makes a mistake because she makes a mistake because she makes a mistake because you can just you can just you can just step in and fix it. step in and fix it. step in and fix it. So let her So let her So let her do it, because that's what do it, because that's what do it, because that's what she's really she's really she's really good at. good at. I want to close with I want to close with one question to one question to learn a little more about you and learn a little more about you and where you are where you are where you are headed. So I'd headed. So I'd headed. So I'd like to hear: when you like to hear: when you like to hear: when you think about the next, think about the next, think about the next, say, two or three years, say, two or three years, say, two or three years, which in the AI ​​world is which in the AI ​​world is which in the AI ​​world is like a decade like a decade ...What are you really ...What are you really ...What are you really passionate about? What are you passionate about? What are you passionate about? What are you building? What building? What building? What are you working on? What are you working on? What are you working on? What inspires you in your life, inspires you in your life, inspires you in your life, especially when you especially when you especially when you think about how AI think about how AI think about how AI will advance it?

will advance it? So, let me start by saying that So, let me start by saying that I am a teacher. I'm not an I am a teacher. I'm not an I am a teacher. I'm not an expert, am I? I expert, am I? I expert, am I? I have no idea. I'm in have no idea. I'm in have no idea. I'm in Oxfordshire. I'm not in Oxfordshire. I'm not in Oxfordshire. I'm not in San Francisco. I don't San Francisco. I don't San Francisco. I don't have access to any have access to any have access to any influential person. No influential person. No influential person. No one whispers in my one whispers in my one whispers in my ear about the future. ear about the future. ear about the future. I actually turn down I actually turn down I actually turn down most most most early early early access offers because I'm access offers because I'm access offers because I'm just not interested. I just not interested. I just not interested. I just want it to just want it to just want it to work now, work now, work now, you know? And so far you know? And so far you know? And so far it's working well for it's working well for it's working well for me, particularly from the me, particularly from the me, particularly from the perspective of my mental perspective of my mental perspective of my mental health. Because you know, health. Because you know, health. Because you know, it's easy to get carried away by what it's easy to get carried away by what it's easy to get carried away by what will happen in will happen in will happen in the future. And your the future. And your the future. And your question was about question was about what will happen in 3 what will happen in 3 what will happen in 3 years, or what years, or what years, or what excites me. excites me. excites me. So, what excites you? So, what excites you? So, what excites you? Yes. I'm excited about Yes. I'm excited about Yes. I'm excited about finding the finding the finding the right way right way to turn my to turn my ideas into work, aren't ideas into work, aren't ideas into work, aren't you? And on the code. Because I you? And on the code. Because I you? And on the code. Because I still don't think I still don't think I still don't think I tried...I think what tried...I think what tried...I think what I'm really interested in right now I'm really interested in right now I'm really interested in right now is is is software factories. software factories. Ways to Ways to automate the automate the automate the start-up and start-up and start-up and self- self- self- improvement of improvement of improvement of codebases over time. I codebases over time. I codebases over time. I think agents are think agents are think agents are not very good at not very good at not very good at thinking thinking thinking ahead about the ahead about the ahead about the codebase right now. And what codebase right now. And what codebase right now. And what I'm excited about right now I'm excited about right now is the transformation of is the transformation of is the transformation of agents into that agents into that agents into that annoying annoying annoying developer on the team developer on the team developer on the team who says, "No, we who says, "No, we who says, "No, we have to do this have to do this have to do this right, not right, not right, not fast, right? We fast, right? We fast, right? We have to do this for the have to do this for the have to do this for the benefit of those who come benefit of those who come benefit of those who come after us." Agents are after us." Agents are after us." Agents are really bad really bad really bad at this right now, and at this right now, and at this right now, and I think there's a way I think there's a way I think there's a way to teach them to do it to teach them to do it to teach them to do it well. That's what well. That's what well. That's what particularly fascinates me. And

particularly fascinates me. And particularly fascinates me. And I think we haven't I think we haven't I think we haven't found the right found the right found the right way to organize way to organize way to organize many agents yet. I many agents yet. I many agents yet. I still think there still think there still think there 's a lot of 's a lot of 's a lot of potential in this, but nothing potential in this, but nothing potential in this, but nothing I've seen has I've seen has I've seen has made me say, made me say, made me say, "Yes, this is the right way to go "Yes, this is the right way to go "Yes, this is the right way to go ." So, I'm also ." So, I'm also ." So, I'm also glad that open glad that open glad that open models are getting models are getting models are getting better. That's another better. That's another better. That's another thing that I'm really thing that I'm really thing that I'm really excited about, because they're excited about, because they're excited about, because they're improving, and improving, and improving, and eventually we'll have an eventually we'll have an eventually we'll have an open model that's as open model that's as open model that's as good as Opus 5.5 good as Opus 5.5 , and I'd be thrilled , and I'd be thrilled , and I'd be thrilled about that. This is about that. This is about that. This is fantastic. I don't fantastic. I don't fantastic. I don't need the models to need the models to need the models to be that be that be that good, right? Because I'm good, right? Because I'm good, right? Because I'm already good enough, I already good enough, I already good enough, I can kind of can kind of can kind of make up for their make up for their make up for their shortcomings. Just having an shortcomings. Just having an shortcomings. Just having an open open open Opus 5.5 level model would be Opus 5.5 level model would be Opus 5.5 level model would be great. And even more than that great. And even more than that great. And even more than that , just , just , just to get more to get more to get more token efficiency token efficiency , to get a good, , to get a good, , to get a good, understandable model that understandable model that understandable model that can quickly produce can quickly produce can quickly produce results. I don't know. results. I don't know. results. I don't know. This is cool. This is an This is cool. This is an This is cool. This is an exciting time. exciting time. exciting time. That's true. That's That's true. That's That's true. That's true. Well, true. Well, true. Well, Matt, thank you for Matt, thank you for Matt, thank you for coming today. coming today. coming today. I really I really I really enjoyed this enjoyed this enjoyed this conversation, and thank you for conversation, and thank you for conversation, and thank you for all the valuable advice all the valuable advice all the valuable advice you you you shared today. Where shared today. Where shared today. Where can people find you can people find you can people find you if they want if they want if they want to learn more and to learn more and to learn more and see what else you see what else you see what else you do?

do? do? And here we are at dot dev—it's the And here we are at dot dev—it's the And here we are at dot dev—it's the same place. I same place. I same place. I have a mailing list there. I am on Twitter. have a mailing list there. I am on Twitter. have a mailing list there. I am on Twitter. I am on YouTube. These are the main I am on YouTube. These are the main I am on YouTube. These are the main three resources. three resources. three resources. Perfectly. Well, Perfectly. Well, Perfectly. Well, thanks again, Matt, thanks again, Matt, thanks again, Matt, hope to hope to hope to see you again soon. See see you again soon. See see you again soon. See you later, friend.