The Koerner Office - Business Ideas and Deep Dives with Chris Koerner - Learn AI Automations Before You’re Replaced by Them ⏐ Ep. #192

Episode Date: July 7, 2025

Check out my newsletter at TKOPOD.com and join my new community at TKOwners.comI sat down with Michael Shimeles to talk about the most approachable path into building with AI. We got into why you don���t need to be technical to start building powerful AI automations, how to sharpen your edge as a non-developer, and why “just building stuff” is the most important thing you can do right now. We covered tools like Claude, ChatGPT, Gemini, Make, n8n, and Gumloop, and even talked about wrapping these workflows in simple web apps. We also explored AI-first product ideas, smart ways to escape the “tech bubble,” and how beautiful design can become your moat.You can find Michael on X at x.com/rasmickyy and check out what he’s working on at rasmic.xyzHere are direct links to all the tools, platforms, and APIs mentioned in the episode:🛠️ AI Models & LLMsClaude (Anthropic / Cloud Code): https://claude.aiChatGPT / OpenAI: https://chat.openai.comGemini (Google): https://gemini.google.comPerplexity AI: https://www.perplexity.aiGrok (xAI / Twitter): https://x.ai⚙️ Automation & No-Code ToolsN8N (Workflow Automation): https://n8n.ioMake.com (formerly Integromat): https://www.make.comZapier: https://zapier.comGumloop: https://gumloop.comReplit (Browser-based IDE & AI dev): https://replit.com💳 APIs & System IntegrationMCP (Multimodal Communication Protocol): https://www.anthropic.com/news/model-context-protocolLinear (Project Management): https://linear.appTimestamps below. Enjoy!---Watch this on YouTube instead here: tkopod.co/p-ytAsk me a question on or off the show here: http://tkopod.co/p-askLearn more about me: http://tkopod.co/p-cjkLearn about my company: http://tkopod.co/p-cofFollow me on Twitter here: http://tkopod.co/p-xFree weekly business ideas newsletter: http://tkopod.co/p-nlShare this podcast: http://tkopod.co/p-allScrape small business data: http://tkopod.co/p-os---00:00 Unlocking Automation for Everyone03:06 The Future of AI Agents05:53 Navigating the Learning Curve of Vibe Coding08:45 Building Useful Automations12:04 Finding Opportunities in Your Industry14:57 The Importance of User Experience17:52 Leveraging AI Tools for Design20:44 Understanding MCP and Its Impact23:42 Embracing Change and Continuous Learning

Are there any low-hanging fruit ways that people could make money with these tools? I'm not a technical person. You're not allowed to say that anymore. You can literally go on YouTube, chatGBT, and learn whatever. We live in a beautiful time. Platforms like NE8 make.com or gumboop. It's a perfect place where an on-coder can experience when it feels like to build something that's actually useful.

I have no doubt you'll be able to build cool stuff. I see a lot of people, Chris, like, get frustrated. Oh, it's not working. I always love this. It's like, give it a few tries. Read up. Ask chat GVT. Hesimists are often right. Optimists make money. I challenge everyone watching. Build at least one after this episode, at least one automation. Use the heck out of all the free trials. And most importantly, have fun learning. When you're having fun, you're not working.

Why don't you start by telling us, like, what type of, like, side projects that you're interested in right now? I'm very much interested in a lot of, like, AI agentic sort of stuff. We're being able to construct agents to do certain tasks. Like something as simple as read my emails and then perform X task. I think that type of stuff interests me more from a learning perspective than from actually building and shipping product. I think if you're going to build something like that, you run the chance to compete with the giant site GVT, chat GVT, Claude and Gemini. And I think the sweet spot in this space we're in is either to build a niche product. like or niche, however you pronounce it, or is to somewhat build something that as the models get better, your product gets better as well, right? I think, you know, early on we saw this a lot and even now we're seeing it happen where people will build features that the models get better and then your feature gets swallowed, right?

Now you can continue to carry on with branding and marketing, right? But I think if you're wanting to build a business, I'm hoping you want to last a little along. It helps to either build your product aligning yourself with the models or building it so niche, so niche that, you know, whatever space you're in, they love you and they love the solution you have. So they care less if the models could do it. So that sort of space interests me. So like to your question, I build a lot of projects, side project in that sphere just so I can know what it takes at what speed I can build. I'm very much at a point where I don't really have any ideas that I'm willing to like shepherd and say this is the one.

So in the meantime, I just sharpen the axe till the idea comes. Yeah. What agents are you seeing that are actually like super useful, super hands off? Because like with my experience with agents, they get like 90% of the way there. But that 10% is so onerous that they can't just run without my input. But what are you seeing today that's like a process, a gum loop integration or anything that you could classify as an agent that really makes your life better or it could make anyone's life better? Yeah. So I'll be a pessimist and optimist.

The pessimist in me will say that we really haven't seen true agentic workflows yet. Everything so much so far has been things being glued together. Right. When you look at NAA or like Zapier or make.com, it's really like a bunch of these tools being glued together, giving them access to these models as tools, which is impressive. Don't give you wrong. I will say the one thing that shocked me in the last couple of days is Claude Code. I know Cloud Code is on the pricier side, but really like, you know, and for the people watching, I, you know, for one month maybe cancel every other subscription and just get one month of Claude Code because it's one of those experiences. where I've been using in the last couple of days,

and it will go on for like 20, 30 minutes, building features, debugging itself, right? Writing documentation. And it felt like everything we've seen and experienced so far was, in my opinion, pretty mediocre. And Claudecote is sort of like, I tweeted upon it recently, it sort of opened my eyes where we really haven't experienced agentic, right?

Because a lot of people have this notion that if it's agentic, it has to be fast. which is not the case, right? If something is agentic, I actually want it to take its time and get it right, versus I have this automation. I have that automation. I'm patching these things together. And sure, they're firing. But the reason why there are mistakes and like you said, there's that 10% where it just can't jump over is because there's no feedback loop for it to reiterate itself.

You have to jump in and either tell it, do it again or change some things up. But with Cloud Code, it's actually the first time I've seen. Oh, it'll keep going until it gets her right. And it will research and it will do all these things that in short, we haven't reached agentic yet. But we're starting to get there. Practical advice. What does it mean for the end user? What it means is it's actually important for people to fill in that 10% gap.

And here's the reason why. I think there is value in being the person who knows how to get the best out of tools. You know, the better driver. Like, we all can drive, right? Some people are better than others, right? We all can build tools and automations. Maybe some can do it better than others, right? So I would say that, yes, we're not there yet, but I would take this time to further build knowledge, further build strategies, further get used to the tools, not because the tools won't change in the next three to four years, but the knowledge that you acquire now will have some benefit as things get better.

Do you think that Claude Code is impressive and just as helpful to a non-technical person or a non-coder as it is for you? There's two sides to this. I think it's impressive to anyone when you see the level of output. But I think the level of admiration will be different, right? There is this love-hate relationship I have with people who vibe code and vibe coding in general. And the reason being is I see a lot of people get frustrated after maybe three hours of vi-haping. coding an app and it doesn't work out and something breaks and they flip out on the tool and they get pissed and it's like welcome to software development they don't have the perspective that you do you know

I mean it takes time right so this is why even when like you know the models were crapping the bed and weren't that good and are just starting to get decently good the simple fact that it would build a simple UI or it would do simple tasks was was was amazing to me because Because, again, the value of software and how much time it takes for someone to architect software, we're entering a time where, you know, code is starting to become cheap, cheap in a sense where it's, for most businesses, it won't be your moat. But I say this to say that a lot of people don't have that understanding of how hard it is to architect a good software. And what I mean by good software is software, if not architected correctly, for the most part, is like milk. it will go bad really quickly. And this is why companies will always have software engineers because even though the software is built well now, maybe packages go out or there's a new way of doing things or there are new features you have to build.

So software is not like why. It does not age well. It ages terribly, right? So if someone has that appreciation and that understanding and they start to vibe code, I think you'll have a better experience versus, oh, I prompted seven times and it doesn't work. work at Tempo. So I have the privilege of seeing people's prompts. People will be like, building me a database. I'm going in circles to say that whether it's Cloud Code or it's tempo, it's all these other tools, I do think the only way people can be successful at vibe coding is if there is at least a fundamental understanding of technical concepts. And those

concepts can help you drive, you know, prompts, decisions, et cetera, et cetera. Yeah. So I think Cloud code is impressive to anyone, but if you're a developer, it's pretty, it's pretty, it's pretty awesome. Would you say you have to be a developer to get the full use out of it? Otherwise, just try Rapplet or something. Here's what I say. I think we're entering a time in a space where I personally, at least maybe I haven't lived long enough. I don't think there is a time in history that we can use to sort of help guide us into what steps are the best. So there's a part of me that's like, try all the tools, all of them. There isn't a single one that you should try. In fact, if I were most people, I would be cycling through all the tools. One month, this tool, the next month, this tool,

even if it's the hard and technical ones, right? I think there is value in learning how these tools work. And the same thing in software development. The learning is usually in the pain where something breaks and you have to figure it out. Right. And before AI, the developer will go to stack overflow or go try to find this one obscure blog written by some guy seven years ago. Like that pain of trying to come up with a solution or understand why something isn't working, although very, very painful is very, very valuable. So I say that to say, as these beautiful vibe coding tools come out, do not rob yourself of the painful learning that you can benefit from exploring every single tool that exists.

So yes, CloudCode is definitely more of a power tool for developers. I still suggest you should check it out. Because what's the worst that can happen? Okay. Tell me about N-N and why you think it is special. The idea of being able to chain certain tasks together, certain actions together, and build workflows is pretty impressive without having to write code. Like platforms like NE8 make.com, like, you know, whichever one floats your boat,

I think are a great place for someone who is an idea of guy or gal should be experimenting, right? Or gum loop. Like there's a lot, there's a lot of great tools out there, right? I think it's a perfect place where a non-coder can start to experience when it feels like to build something that's actually useful. So if I was a non-developer and I've thought about this for a long time, a lot of the stuff that I would do is I would play with a lot of the tools, like I mentioned, but I would also start to automate even the small things in my life, right? It doesn't have to be perfect. I think you rob yourself of having to, like not everything has to be a SaaS. Maybe there's some things you build for yourself, right?

And in doing this, you prepare yourself to be that person who builds something that's pretty awesome, right? So I challenge everyone watching, build at least one after this episode, at least one automation that does something that you do on a regular basis, right? Let's say you get a lot of spam emails, right? Maybe something that reads those and you use GVT to respond in a hilarious troll way. I don't know what it is, whatever it is. Just something to learn. Yeah, just something to learn. I think there is value in learning.

and oftentimes especially with like Twitter, every one is perfect, everything is perfect, people don't make mistakes. At least that's how it looks like when you're looking at other people's stuff. But I really challenge you to just build, have fun,

enjoy the process, break things, learn, enjoy the pain. But that's what it's like being a developer, right? So I think there are certain things that even as the models get perfect, they're not going to read your brain

unless you put the chip in and I don't think I'll do that. But I say that all to say, embrace it, right? embrace it and just have fun. And if it's fun that you're having, it doesn't become stressful. Because I see a lot of people, Chris, like get frustrated.

Oh, it's not working. Oh, is that this? It's like tools don't get built in a day. At least something valuable, right? If you're building a language, I understand. But, you know, most people don't even know the different types of databases and are frustrated when after one prompt, the front end's connection or the back end's connection to the database doesn't work.

Yeah. Give it a few tries. Try again. Try tomorrow. maybe read up. Ask Chad GVT. One thing that I do is for every place I get stuck at, just because I want to learn from it, I will literally either screenshot, whatever era I've experienced or where I've got stuck at, and I'll chat, GBT. What happened and how do I learn from this? I think just investing in these little small actions

prepares you for tomorrow's world because I'm one of those people that doesn't think, like I'm not an AI is going to replace every one type of person. I am a person who says AI is to change things. And the people who change with the times and update themselves with the times will be the ones who reap their awards. There's a same. Pessimists are often right. Optimists make money.

Yeah. Yep. I love that. I love that sting. What is the coolest automation you've built and are you willing to show us how it works? Yeah. So my favorite one, it was an AI pizza ordering agent. Oh, that's cool.

Basically, it went to my favorite spot, which is Dominole's.com. Domino's.com. Domino's.com. Domino's.com. Domino's the best beats of place in the world. So what is it essentially did was I, the user, had a chatbot interface. So I also built a web app for this. So it wasn't just an automation. And I'll kind of explain how you can build both. There was a chatbot interface. And this chatbot interface, I'll use Circle. had an AI agent I built.

I use agents loosely, but you get what I'm trying to say. Yeah. And this agent interfaced with Domino's API. Now, I will say the reason why I didn't publish this or I don't maintain it is because the API that it used to interact with Domino's was like not an official API. It was like some guy who built it six years ago. and like there's no really secure way to handle payment. So it's a little suss, which is why I only built it for myself.

But basically the way it worked is, user would come on the site. The AI agent would notice that they're on the site, would say, oh, would you like to order pizza? And then it would give you four options because that's how I configured it. And then it would go back and forth. So we would have this conversation with you where it asks you a question. You give it an answer. That answer becomes, gets stored in a DB.

and then it keeps going back and forth. And then once everything that was needed to make the order was set, once the AI agent had everything it needed, it would then call the Domino's API, make a payment, and all that stuff. Here's where things get interesting and maybe we'll have this as homework for people is I think NAA web hooks. Yeah, so N8, and I'm pretty sure Gumloom, that's what's hot in the streets now, I believe. But with any aid, they have what you call web hooks.

And basically, what's awesome about this is you can connect whatever automation you've built. Like, let's say I have an automation that every time I get an email, I'm a YouTuber, I get people email me if I need someone who make my thumbnail. So maybe I have an automation that, you know, read all my emails. and if someone mentions thumbnails ask how much, right? Something simple, right? I don't want to go back and forth. I just want to know how much, right? If they're email me, they're probably sending me their portfolio, what they have,

ask for how much. And then you can get fancy where you can get it to send you like a message to WhatsApp, telegram and all this stuff, right? But what's interesting with WebHooks is I can connect this automation with a web app. And that's when it becomes interesting. So let's say you have an automation where a user just needs to give you their email. And then what you will do is you will scrape the internet for a specific shoe.

And if it goes below a certain price, then it sends them a notification. You can build that automation in NAA, right? But how do you package that? How do you sell that? Or how do you build an interface? that makes it very, very simple for someone to use that automation, right? Web hooks is where it comes in handy. So, again, I'll have it highlighted here.

You can literally just search and A and 8 and then web hooks. The first doc will take you there. Again, will you get this right the first time? Is this painful? Will you have to vibe code a few things if you don't know how to code 100%. But if you do this, most people don't know how to do this. I have no doubt you'll be able to build cool stuff. So I'm very much interested in how do I turn these cool automations or really agents into something that a user can use, right?

So this is why you can maybe call this vibe coding plus vibe automating. So vibe coding plus vibe automating. You take something you built in NA8. I will admit as a developer, I like building. these myself, right? But I still learn how to use the tools because, again, I want to be up to date with the times. But if you're someone who doesn't know how to code, you take some of these automations and then you take a Vibode tool, I'm biased, I'll say tempo or any other Vibco tool of your choice, I would really learn in building something combining these two, right? I think

you start to build applications that are more useful than your standard landing page, but you also start to see how these different components work, right, how a website works, right? Like, you don't have to be a developer to know that, okay, every, like, functional app has a front end and a backend, right? The front end is the user facing side. The backend handles all API calls. You know, if a website has username, password, and all that stuff, that's authentication. If I'm doing something on a website, I log out and I come back in and information's there.

That means a website has a database, right? These things are not, shouldn't be guarded as, oh, only technical people should know this. I think it's a very easy excuse. I think you can know base level how these things work. And I guarantee anyone who knows even like, and this is why product managers, designers, they cook with these type of tools because they have technical knowledge. They're not technical people, but they have technical knowledge. So when you combine that knowledge with these AI tools, whether it be automation, vibe coding, cloud code,

whatever it is, almost always, those are the people who end up having the most fun in building the coolest things. The non-technical people, the product managers. Well, people with kind of a base level technical knowledge, but not coders. I say non-technical people in technical organizations or environments. Yeah. Okay. When it comes to building your average automation like this, what tools do you like to use

for what types of tasks? And you know what, let's expand that out. What LLMs, what tools are your go-to tools for certain tasks? Yeah. So I will draw a line. I have, let's say, a spectrum, right? Let's say from user-facing cool, vivy to genius, code, difficult. Right?

Let's say this is a spectrum, right? Yep. If I'm building anything where code is being generated or I feel as if the task is very, like, require some sort of calculation or whatever, anything difficult, I'm always going with Claude. Okay. Right. And then on the opposite spectrum, let's say I want to build like a nice image generator, right, for people to generate cartoons or I want to build a poem generator.

I want to build an agent that talks in slang, right? Almost always open AI. And if you use open AI enough, like, I don't want to open my personal chat GBT because it show my history and stuff. But my chat GVT starts to talk like me. Like it will say, brough, but B you are, UH, UH because I talk like that, right? So like the model is sort of designed to be in a very psychopathic and weird way, like almost personable.

Yeah. So it almost becomes my go-to for like, I don't even want to be to that guy, be that guy. If I'm building anything, B-to-C. And I wanted to have a personality. I'll always go for something like Open AI. And for anything, again, agentic, writing code, complex stuff. Binary. Binary outputs. I'm going for something Claude. And then Gemini, honestly, is like an all-star where. Gemini can sort of be both, right? And there is a use case to Gemini where like Gemini is also the most cost effective. Where and maybe we can have another graph here just to paint this picture. One of the things that, especially if you plan on building something that's going to be used a lot or scaled is something you want to think about early on. If we were to compare price, quad is up here. Gemini is, you know, very cheap.

And then Open AI, depending on the model, is very right there. So there are a lot of things I would take into consideration. In the initial building stage, I'm not really worried about cost and neither are you. So I wouldn't really think too much about that. I think you should only worry about costs when, you know, you've made more than a million dollars. Then it makes sense. Yeah. But until then, it's just a matter of what type of app are you like building anything.

time I'm building anything user facing, building a basic chat bot, I'm always going for open AI. But if I'm building any coding assistance, anything that I feel like requires more agentic, especially with Claude 4.0, I'm going with Cloud. Right. And then Gemini is like an All Star where like I, you can pretty much use it anywhere. They have an image generator. They also have 2.5 pro, which is a great coding assistant. It's not as good as Claude 4.0, but it's pretty up there. So this is how like I sort of played myself. Now, for someone who is a beginner at, you know, let's hear building some sort of automation you want to pick an agent, just go with Open AI. Just go with Open AI. Almost always, open AI should be enough for you. And I would go Open AI and Gemini. Claude is on the more

expensive side. I don't want people to lose money. So I would almost go, always go with Open AI. And if you're really like cost is a big thing, then Gemini. So that's kind of how I look at these models. This is a spectrum I think of in my mind. Where do perplexity and GROC come into play for you, if at all? I'll be honest. Like these models, like these three just do it well for me. Like I have not used GROC in terms of the like the API.

Now I've used GROC like to like answer tweets to describe tweets. That feature is great. But it's never been like compelling enough. for like I wanted to use it like I wanted to use the model. Proplexity, if I'm going to be honest, open AI's like deep research and search is just such a great product that I used to use perplexity, but not anymore. So like I'm satisfied with these tools, right? There isn't something those tools do or those models do that is not here.

Now if one of them launches a new model and it's exceptional and it's well and it does something better than these guys, guess what? I'll move over, right? Part of the game is being able to know, okay, what's out, what's popping, which one's doing great, and then being able to move over, right? Just because this is my set today doesn't mean literally tomorrow won't change because I will tell you before Cloud 4.0, I was literally using just Open AI Gemini. Cloud 3.7 was not good in my opinion and it just felt like a waste of money. So I literally stopped using Claude and it was just Gemini Open Aad. 4.0 dropped and I'm completely on the Cloud train now, right?

So part of the game is being flexible and malleable to like all the new tools coming out. Are there any like low hanging fruit ways that people could make money with these tools, whether that's building a wrapper or starting as like an agency or services business, AI implementation, AI automation implementation business? What's top of mind for you nowadays? Yeah. Like actually I had an interesting conversation at the office today. And someone was like we were talking about like some of the new.

new apps that have come out. And myself and a fellow engineer were like, some of these things are so easy to build. Again, talking from an engineering mind. And we were shocked at one of the companies we were looking at raised so much money. Like I think upwards to like $10 million for something that honestly would have probably taken us maybe two to three days to build, right? And someone who was there said, you know, oftentimes because we're all in this tech bubble, we feel like some of the stuff that's normal to us isn't impressive to somebody else. isn't valuable to somebody else. And sometimes it just takes poppy out of that bubble.

So I say that to say, the best opportunities, at least that I think right now, are those where you look outside of this tech bubble, right? Because in this tech bubble, look who you're competing with, right? There are so many bright minds and so many engineers and not technical people and technical people. This tech bubble we're in is not the best place to start, especially if you're a beginner, right? So if you are not a teenager and you're an adult, you've had some experience or you work a job, you've been part of some industry, I would literally start there. What are some things in your industry at your job, at your office, or your company? What are some workflows?

What are some tasks? What are some things that you dread? You would be surprised how receptive people would be if you solved problems that you genuinely understood. Right. Let's say, for example, Chris, I'll throw out a free idea. I really wanted to build an AI bookkeeping app. Tax season is hell. Keeping up with receipts is annoying.

I download all these apps, quickbooks, zero, whatever. They mess everything up. Everything is not organized well. My rent transactions are categorized as business. It's a mess, right? So there's a lot of manual and tedious work. And there was a part that's like that I wish this app existed and maybe I should build it. But I didn't build it.

The reason why I didn't build it. I'm not a bookkeeper. I'm not a accountant. So I would build something surface level. And sure, it would work for me. But imagine an actual bookkeeper. Imagine an actual accountant. Someone who's in the trenches, who knows the pain points, who knows the long task.

I think if you are that person, and I'm not saying you have to be professional. I'm saying you just have to have a job, literally. And I'll talk to the folks who don't have jobs to the young ins. for those people, that is the best place you should look. It's terrible, especially as a beginner, to be like, what's the next, like, can I maybe build a model or try to do like something egregious and insane that's in this tech bubble that there are thousands of people who are in this space daily building it.

But if you're so non-technical, who is having fun using these apps, who's learned how to build, you know, something cool with these tools, start from industry in, right? number one, the best problems to solve are your own. Number two, the best problems are to solve are the ones that you're in that industry. You're in that business. You're in that. Like, that to me is the gold mine. Because you know the customer.

You have the customer. You work there. You are there. You know the problems and you even know how to communicate, right? Because building it is part one. Being able to market it is a whole other behemoth on its own. And you can do that well if you're in that industry. Right.

So I would start there. My place, my job. my people, whatever that situation is that you're familiar with, solving those problems, solving. For example, I have a friend who works at a pharmacy and he's having fun building things for pharmacies. And it's not because he's a super technical genius, but it's because he's worked in pharmacies. He knows how annoying billing is. He knows how annoying, like the software is that where they register medication and all that stuff. He also knows all the compliance. that's required, right? And these are all boring problems for someone who's not in that space to

solve, but for someone who's in the tech space to solve, if you're stuck in that tech bubble, what does solving pharmacies problems sound like? It sounds boring, but that's where the money is. If you can solve the problems for a specific industry, especially your partner, you're golden. I love it. I mean, even if you're, everyone's in an industry, but you're having a hard time thinking of problems in your industry. Ask Chad GPD, like, hey, I'm a pharmacist. I want to build an app for other pharmacists? What are some common pain points? I would use deep research for this. I would say, I want you to go deep. Like go to Reddit, go to message boards, go to social media, find all the gripes, all the negative tweets, all the complaints about the pharmaceutical industry and list them

all out for me. It lists out 20. Okay, take the seven that you think are most relevant. Pace those down and say, okay, now I want 20 more that are a lot more like these seven. So you can just keep refining it. Pretty soon you'll have 30 to 50 problems and all of them will ring true to you as someone. in the industry and then you can say, okay, I'm going to use, I'm going to take the top three and build rappers from these three and distribute them to my friends and coworkers just to test and play around with and see what feedback I get and then iterate from there. Right. Yeah, 100%. And I think like you mentioned it, like going back to actually using AI to do their research, right? I've actually done something similar where I go to church every Sunday.

I serve at a church. I teach at a church. And, you know, there are certain things I've realized that I'm like, Oh, like I live in Canada as well. So, for example, payments for a lot of charities that the softwares are either terrible, the percentage they charge is very, very high. And like you can't just use like Stripe. You have to build a website and stuff like that. So I was like, huh, I'm going to build a payment solution for my church. And I did that.

If I wanted to turn it to a business, I could. I don't have time. But if I wanted to, I could. But you see, and I almost guarantee if I wanted to build that business, it would do so well because it's a problem I experienced and I know the industry experience. I built the solution and I'll price it much fair than the competition. It's just a perfect formula. So I think starting where you are and speaking to your people and when I see your people,

I'm not talking about race, creed or anything like that, but I'm talking about your industry, in your business, in your line of work, I am telling you the things that if you think in a tech bubble mindset, you'll always think, ah, this is simple. this is boring. But at the end of the day, first principles, people solve, people pay to get their problems solved. As long as you're entertaining people or solving their problems, you should be getting paid, right? So that's how I think of things. Don't be too tech inflated. I think sticking it to your industry and the problems in your space and solving that is a great way to start. I like the back to

your bookkeeping, your AI app for bookkeeping. It reminds me of Cal AI. And for those listening, Cal AI is an app that basically said, my fitness pal has hundreds of millions of users. It's a nine figure company or it has $100 million plus of revenue. We're just going to do that except AI first, right? And that one sentence, that one liner, that model is like that, but AI first, right? Instagram was like, oh, Facebook, that's cool. How about mobile first Facebook? Pictures forward, mobile first.

We're not even going to have a desktop site at all. We don't even worry about that. And all these other apps in like the late 2000s, early 2010s, their whole model was just, oh, we're going to take that website and do mobile first. An actual desktop version of this may never happen. And that's fine. So now it's just like, oh, we're just going to do that except AI first. AI first and foremost, AI foundational.

So that could be QuickBooks except AI first. That could be anything except AI first. And QuickBooks, My Fitness pal and all of those, they're going to integrate AI. But they will never be AI first. They'll never be built with AI first and foremost, right? And one thing I would add to that too is like a lot of the like the new apps like CalAI and stuff that you mentioned, they have one big differentiator. And this is like I plan on making a video on this.

And this is going to be a moat for a lot of people. And that's user experience, how your app looks and feels. I'm notorious for two things. If you know me personally, my like IDE, like the where my, where I write my code, I change the theme every like two, three days. Browsers, I change every five or six. And the reason being is, I love

the place that I'm working or the apps that I'm using or the sites that I'm using. Whatever I'm using, I want it to look and feel nice. I say that to say that most people, most people, most people, most people, most people. If you focus on crafting a beautiful and elegant user experience, that most of the time is almost enough

to do well. You have, like, for example, there are ticketing softwares, like, where you can, like, use, like, you know, you can, like, let's see you're running an engineering team and you need to track, like, tickets and who's building what, and you assign tasks and stuff. There's stuff like Jira, for example. Jira has been around for, like, maybe as long as I have. And it's an ugly app. Yeah.

Yeah. It's ugly. And, but then you have something like linear, which. basically is Jira. But look at that. Which one? Gira except pretty. That's their business.

Pretty fast, right? Like they're very thoughtful with every single thing that they do. And I think in a world where everyone is saying, oh, AI is coming for the devs. AI's coming for this. The people who care about these little things, right, how the app is crafted, how the user experiences, the features, what makes things pop? what makes someone want to just use their app, even when they don't need to, I think those people win big.

Yeah. UX becomes a moat. I call it the sauce. I think UX becomes a mode where if you're someone and homework for anyone who wants to do this is I would collect sites and apps that you wish you built. Not because they made money. We'll worry about the money stuff later. Yeah.

Yeah. But because it made you feel a special type of way. For me, like I use linear. And every time I use it, I'm like, man, I wish I'd like. Yeah. Like, yeah, I wish I could put my face on this and say, I built this. And that's how you know a software is good. So I think being able to craft and develop a taste for how user experiences should be, how apps should be, how sites should be.

Even as a non-technical person can be very fruitful. I actually think, Chris, there's going to be a time where even though the models are going to get better, that, and this has happened for years, but I think it's going to happen even more, that a lot of non-technical people are going to hire, are going to partner. up with technical people, right? I think we're still going to enter, we're still going to be in a world where you like devs are needed and will help supercharge things. But I think it comes to a point where the non-technical person will start to be valuable to if they take these things into consideration, right? But if you just, you go to some app and you're like, build me. No plan, no idea, no understanding. Right. And again, this is not an excuse. The excuse isn't, oh, I'm not technical person. You're not allowed to say that anymore. You can literally go on YouTube and

learn whatever. You can literally go chat gvt and learn whatever. So say that to say we live in a beautiful time. Well, like they say if you build it, they will come. And I think you would say, if you build it beautifully, they will come. Right. Like beautiful apps. They do. I mean, look at it this way. Let's say you're looking for a co-working space in Canada. You go tour, two of them. You go into one and it's like a little dirty. It's kind of gross. And there's no artwork on the walls. And it's just bare bones.

You could tell that they cheaped out. They didn't put much thought into it. They just wanted to get the max value per square foot as possible. And they did some things to kind of look nice, but you could just tell, this is nice. And in a vacuum, at this were my only option. I'd probably use this because I need a co-working space. And then you go to the next space next door. And it's gorgeous.

And there's indoor plants. And there's furniture from restoration hardware. You're like, that's a $12,000 couch. I feel good in here. This is good energy. Like you're going to not just choose that one, but you'll probably pay twice as much to use that one. So everything I'm outlining is obvious.

Like everyone would rather choose that one. But we don't treat our software like that. We just throw out something, our MVP. And then if it works, then we leave it like that forever. Because if it's not broke, no fix it. But like look at Airbnb.

They were designed first from day one. And it's a hundred billion dollar company. Like that was their whole strategy. design, right? We're going to have a beautiful design. And I think, I think it matters. I think it matters more than ever. And like, unless you are building something very innovative and game changing, maybe you can get away with having ugly UI, but we're entering a time where the expectations high, like, if I see a site and it looks like it was made in 2005 unless it's a government site or whatever like or like some like you know farm like whatever like a hospital site or whatever there is no excuse because they

they have like compliance and all that stuff and or banking whatever that's fine but any business any SaaS any startup like design is super important so make sure that you know you develop that taste in that craft. And it comes from, you know, looking at sites that you like, for example, I can go on the gum loop side and maybe five, you know, I like the little diagram they have here. And another thing you can do is actually go to popular like frameworks. Like, for example, there's magic UI. Like this is like a component library. Like you can see these different designs. Yeah, sure, they're meant for developers. But again, just being able to develop that taste of like, Yeah.

This is a nice animation, right? And then you have obviously the Crem de la Crem Apple. Like what's beautiful about Apple? Why? Like, what's their style? Oh, you see they have a specifics font. You see that they're spacing and the letters and mental grid. Like I actually spend a considerable amount of time just doing this, admiring these beautiful design, elegant user experiences.

There's another one. This is a crypto startup, but nonetheless, they're again. Like a lot of people hate. I bet if you went to the site and you just used it, you'd feel a certain type of way. Right. So I think these things become a moat and you don't have to be a technical person to develop this taste. And if you're betting on the fact that the models are going to get so good that it'll be

able to build stuff for you, I think being able to develop the part of you that can help guide the model to build stuff like this is important. Right. And that can do time. That comes through building. stuff and breaking it that comes through like you know testing at all these tools and i don't hear anyone speaking this but i highly suggest make some friends that are developers because i know everyone's talking about oh developers are going to get replaced but i'm telling you you have you a couple of

developer friends that you keep around you buy them pizza everyone in a while it once in a while i am telling you it will yield results not only from knowledge but also from like you know maybe you partner up together and build stuff together or maybe they know of tools that you don't know or maybe they even answer or help things like the amount of like people that I know who are like cousin of mine was vibe coding and the tool he was using I'm just couldn't get the API stuff down and he like spent two three days so guess what I did Chris I hopped and not fix the firm is a 30 second fix for me it's easy right I think and then he continued to build his app right so make some developer friends I think it'll be very helpful that's a good idea when it comes to beautiful design you know of any

AI tools out there where you can say, hey, copy and paste this website or copy and paste this app. Unfortunately, many of them will not give you a one-to-one copy. You will have to, this is a work fly. I would say if I were to do that. So step one is screenshot, right? Step two is obviously feed screenshot to AI. Which AI? Again, whatever Vibeco tool you want to use. It honestly at this point in terms of design, it doesn't really matter. You start to see that certain models have like certain design tastes or like lean certain ways. That's kind of up to you to decide, right? All of them.

But you're going to get a result. And the result is maybe 40% of what it actually looks like. The way you get the 40% to 100 is by going to. through every section of the site that doesn't look pixel perfect and telling it. So say, for example, I take a screenshot here, right? Let's say I take a screenshot here and let me just pull the screenshot here. And I dropped this in the model. I tell them, make it look like this.

I almost can guarantee you it will not. It will get the layout down. It will get product, resources, pricing, customer blog content. contact down and get the buttons down we'll get the text down but it won't get it'll mess up someplace so what my job is now is to go section by section and start to tell it to fix it so for example something that you might not have noticed here look how rounded this button is this is not a square button right has a little bit of roundingness to it so what you're going to do is you're going to you can screenshot this again and then upload it or you could just tell it the buttons aren't the

same. Please round it more. So that's one part you can fix. And then you can maybe see that you notice how there's a grade like it gets darker here. Right. It's not just white. Like it's white and turns the grade and starts to get darker and darker. Right. You start you have to pinpoint each section where it does not look perfect and then make it perfect by telling it. Now here's advice that might sound counterintuitive. But hear me out. If you're building something, and I used to say this and I'll explain why I used to say this but I'm going to say the opposite

if you're building something I think the landing page should be the last thing you should focus on I think design should have its own time if you're thinking about design and features at the same time especially as a non-technical person I wouldn't do it I think design deserves its own time I used to tell people build a landing page first and I think for very very beginner people I would say that

And the reason why I say that is the dopamine, right? It's momentum. Yeah, it's moment. Right. It's like you get the landing page done and it just feels great. It's exciting. Like you have this dopamine rush. It helps you, you know, to keep going.

But if you're on the discipline side, you've done this before, this isn't your first rodeo. I would start to build, you know, the stuff that you're actually trying to build. And then you could put the landing page on top and focus on design, right? But if you're a complete beginner, you know, start with your landing page. get the dopamine hit, get excited. But I would not start with the design first unless I have no idea what I'm building. And I just want to have a nice dining page. Yeah, beautiful.

Two questions. Number one, can you explain in two sentences what MCP is and why people should care about it? The simple phrase I've been hearing is the USBC for LLMs. Okay. Meaning like the new iPhones have USBC, my MacBook has USBC, That person's Android has USBC, right? We all can use the same charger. We all can share the same SSD drive, right?

It makes it easy for all these different devices to use that external functionality. In the same way, what MCP allows you is it allows for LLMs to easily access external services, right? So, for example, linear, right? Linear is a ticketing tool, right? Let's say I have an LLM here. It could be Claude. It could be whatever. And I have linear.

I have linear over here. Now, if I want my LLM to have access to linear without MCP, let's just say MCP doesn't exist, I would have to go to Linear's API documentation, right? I would have to learn how their documentation works. And then I would have to basically write some code. and I have to, I'm assuming I have some back in, I'd have to write some code to connect to linear's API. Not difficult, but what happens when you have 20 different things you want to connect with?

There's many different APIs, documentation, not everything is the same. It gets very annoying. MCP is a standard where third party services like linear will follow a specific rule set on how to build their MCP server. And the LLMs, aka the MCP client, knows exactly how to connect to these servers, right? It's like the USB port. So the LLMs have the USB port, linear and all these third-party tools are building their product with a USBC in it.

So it's easy to connect and plug in, right? And for most people, this might not sound like a big deal because, again, there are still LLMs that connect to APIs, but from a production standpoint, from building out these robust, like if we ever wanted a Jarvis, right, like an Iron Man, where you tell the assistant, oh, fetch my email, contact this person, make sure to close this ticket and all that stuff. To build it without something like MCP, it's a very robust job. Now, MCP isn't perfect. There's still a lot of caveats and stuff like the Anthropic team and stuff are figuring you out.

But it's to say that it's a standard for LLMs. And it means that we're going to be living in a world where the LLMs will essentially have access to anything that has an MCP server, which at this point is almost every third party SaaS service out there. Okay. Okay. Last question. Is it worth someone trying to learn how to use MCPs and N-N or 80-20 of it is learn how to use Lindy and Gumloop and Zapier and you're good to go? Or do you think it's worth taking that extra time? Okay, so my advice is different for different people.

If you don't have a lot of time on your hands and you maybe have one or two hours a day, I would just be an avid user of these LLMs, right? Use chat, GVT. Before you Google it, maybe chat GVT at first, right? Get into that habit of using those tools. But if you are someone, like maybe you don't have four kids and you don't have this big family that you have to take care of. I understand not everyone has a time. so I want to be understanding that.

But if you have all the time, you work 9 to 5 and after 5, you're free, use everything. Learn everything. Maybe you'll like something. Maybe you won't, right? Just listening to what I think you should learn is pointless because literally as soon as this episode airs, tomorrow, one of the tools might launch a crazy update and maybe everything I've said changes. The one thing that I want people to take from this is we're entering a time, we're entering a new, I don't want to sound cliche and say, industrial.

Industrial Revolution, but we're entering a time where things are changing. And being part of the change means learning everything. Some of it will fail. Most of it will fail. Some of those succeed. But the knowledge, the experience, they can never take that away from you. So use all the tools. Learn all the tools.

You know, use the heck out of all the free trials. A lot of these tools have free trials. Take advantage of all of that. And most importantly, have fun learning. I think when you're having fun, you're not working. We're having fun, it's not work. And when it's not work, you can do it continuously. Right. So that's the biggest advice I would give to people.

Have fun building using these tools. Man, that was beautiful. Thank you, Mike. I appreciate that. It was amazing. Where can we find you? Yeah, so I'm on X at Ras, Mickey. I can maybe pop it up here. You can find me here. I troll every now and then. I also have a YouTube channel where I talk about dev, non-deaf stuff. Obviously, Chris, I really appreciate you bringing me on here.

It was a pleasure and hope people learn something. Of course. Thank you. All right, what did you think? Please share it with a friend and we'll see you next time on The Kerner Office.