Ep 102 - A Simple Explanation of Agents and Vibe Coding with Greg Howe

Summary

---

The ChatGPT Experiment – Ep 102: A Simple Explanation of Agents and Vibe Coding with Greg Howe

Date: March 21, 2026
Host: Cary Weston
Guest: Greg Howe (Founder, Gimme Info LLC)

---

Episode Overview

In this episode, Cary Weston brings Greg Howe back to demystify two important yet often-misunderstood AI concepts: agents and vibe coding. Aimed at curious beginners, Cary and Greg break down these terms using real-life analogies and plain language, offering listeners foundational understanding without deep technical jargon. The conversation is focused on empowering professionals and enthusiasts to feel comfortable with these concepts, know when and why to use them, and understand their growing role in practical AI solutions.

---

Key Discussion Points & Insights

1. Why These Topics? (03:31–05:30)

• Cary explains why terms like “agents” and “vibe coding” often confuse people—even those who use AI tools regularly.
• Many newcomers feel pressured to already understand AI terminology.
• The aim: offer “water cooler” conversational clarity, not a technical tutorial.

Quote:
“There's a lot of folks that use words and terms that we're thinking we're supposed to know... but oftentimes we don't.” — Cary Weston (04:10)

---

2. What Is an “Agent”? (06:39–16:00)

a. Simple Definition (07:32–08:52)

• Greg’s 101 Explanation:
  An agent is an AI 'employee' (or ‘contractor’) with a singular job. It takes an input, does its job, and gives an output.
• Agents can “talk” to each other—one agent’s output can feed into another.

Quote:
“The idea of an agent is kind of what you would consider to be an AI internal. An agent has a job and you go say, here, here's some input. Go do your job and produce some output and let me know when you're done.” — Greg Howe (07:35)

b. Agent vs. Custom GPT (09:14–10:04)

• Custom GPTs are ChatGPT instances with specific instructions—essentially, focused, purpose-driven agents.
• Agents: job-focused, input→process→output.

Quote:
“What makes it an agent is that it has an input, it has a job to do, and it has an output.” — Greg Howe (09:35)

c. “Contractor” Analogy & Decision-Making (10:12–11:35)

• Agents can apply logic and make simple decisions (e.g., skip a step if not required).
• Like a plumber with tools: you state a problem, they determine how to fix it.

d. Where Do You Find Agents? (12:09–13:17)

• No “agent marketplace” yet—most agents today are built for specific needs.
• Start from your business problem, not from the tech.

e. Step-by-Step Guidance for Beginners (14:55–19:27)

• Open ChatGPT (or similar tool), explain your problem in plain language, and ask if an agent would help—request consultation, not just code or tool suggestions.
• Focus on the business use-case first before diving into AI specifics.

f. Summary Checklist for Beginners (18:23–19:27)

• Use AI tools as consultants.
• “Find your one thing”—start with a single business pain point.
• Agents can be stacked/reused for different problems.

Quote:
“Once you have an agent that does a good job... it can also be used for other problems that you might want to solve. So these agents are reusable. You stack them differently.” — Greg Howe (19:53)

---

3. What Is “Vibe Coding”? (21:03–35:21)

a. What Does "Vibe Coding" Mean? (21:03–22:32)

• Vibe coding: describing what you want in plain English, letting AI write the code, and then iterating based on your feedback.
• Works for both visual and functional programming.

Quote:
“You describe what you want in plain English, AI writes the code, and then it becomes this iterative process where you say you want to change something and it changes it.” — Greg Howe (21:03)

b. Real-World Examples (22:13–25:07)

• Greg built an entire SaaS product via vibe coding—never wrote a line of code, just iteratively instructed the AI.
• Monday.com Case: Developers cloned Monday.com’s core features in about an hour by describing the site to AI.

Quote:
“I never wrote a single line of code. The entire thing was vibe code. And it works beautifully and it's better than I would have written it.” — Greg Howe (22:13)

c. Practical Application for Small Businesses (24:23–25:07)

• Even non-tech professionals could build custom solutions by conversing with AI instead of paying monthly SaaS fees.
• Caveat: start with solving one pain point at a time.

---

4. Caveats and Limitations of Vibe Coding (25:07–29:29)

a. Manage Expectations

• Don’t expect perfection in the first iteration—AI code/output quality varies.
• Save work in pieces, backtrack as needed.
• AI can “forget” earlier parts of conversations in long sessions (“the backpack gets full”).

Quote:
“Understand that what it generates can sometimes be crap. It shouldn’t turn you away from it... just say to it, that was crap. Do it. I want this to be more professional.” — Greg Howe (25:18)

b. Don’t Be Afraid to Direct the AI

• AI won’t push back if your instructions are unclear—it does what it’s told, sometimes incorrectly if you’re vague.
• Give visual/business context, examples, and feedback.

---

5. Universal ChatGPT Principles in Practice (29:41–34:35)

• Role-play with AI tools: assign them roles (developer, architect, advisor), have them ask clarifying questions, and let them interview you to understand needs.
• Build prototypes and have AI produce “blueprints”—even outside of software (e.g., house plans).

Quote:
“You want to tie in the other part. Create an agent that is your architect. Create an agent that is your senior developer... that doesn’t build code for you, but tells you here are the gaps.” — Greg Howe (33:06)

---

Notable Quotes & Memorable Moments

• On Agent Simplicity (07:35):
  “An agent has a job and you go say, here, here’s some input. Go do your job and produce some output and let me know when you’re done.” — Greg Howe

• On Vibe Coding (21:03):
  “You describe what you want in plain English, AI writes the code, and then it becomes this iterative process where you say you want to change something and it changes it.”

• Caveat for Beginners (25:18):
  “Understand that what it generates can sometimes be crap. It shouldn’t turn you away from it... That’s the whole vibe aspect.”

• AI as a Business Consultant (13:17):
  “If an agent makes sense to solve the problem at hand, then you would have to go use... whatever tool you like, and start saying, hey, I’m looking to create an agent that can do this task for me.”

---

Timestamps for Key Segments

• 03:31 – Introduction of confusion around agents and vibe coding
• 05:30 – Greg’s introduction and approach to AI for small businesses
• 07:32 – 101 definition and analogy for agents
• 09:14 – Agents vs. Custom GPTs
• 10:12 – Difference between "tool" and "agent" (contractor analogy)
• 12:09 – Where agents come from, building your own
• 14:55 – How to approach AI tools as a beginner
• 18:23 – Checklist summary for identifying/using agents
• 21:03 – What is vibe coding
• 22:13 – Greg’s SaaS built via vibe coding; the Monday.com example
• 25:18 – Critical caveats for beginners using vibe coding
• 29:41 – Leveraging core ChatGPT practices for agents/vibe coding
• 33:05 – Role-playing and productive AI consultation
• 36:17 – Greg’s contact details and his experience with vibe coding

---

Conclusion & Takeaways

• Agents: Think of AI agents as single-task contractors. Focus on defining the problem; let them do the task—and perhaps talk to each other. Build or reuse as needed.
• Vibe Coding: Use plain language to describe needs and iteratively refine the AI’s results. Start small and expect several rounds of feedback.
• For Beginners: Don’t start with AI—start with your business problem. Have simple, back-and-forth conversations with AI tools. Use analogies and role-play to structure your exploration.
• Final Thought: Stay curious—comfort and capability with AI grows out of small, practical steps rather than trying to understand or implement everything at once.

---

Connect with Greg Howe

• LinkedIn: Greg Howe / Gimme Info
• Email: [e-mail]
• Website: gimmee.info

---

“Get your hands dirty, but understand it’s just like the contractor mentality: there’s a lot you can do, but sometimes you need a professional.” — Greg Howe (37:03)

---

Transcript

[A]

Security program on spreadsheets, New regulations piling up and audit dread. It's time for Vanta. Vanta automates security and compliance, brings evidence into one place and cuts audit prep by 82%. Less manual work, clearer visibility, faster deals, zero chaos. Call it compliance or call it compliance.

[B]

Get it?

[A]

Join the 15,000 companies using Vanta to prove trust. Go to vanta.com calm hey, it's Kerry. Hey.

[B]

This week on the show, my goal is to help you understand a couple terms that get used a lot, but really not explained a lot. And that's agents and vibe coding. And to do that, I'm bringing my friend Greg Howe back to the show to talk about it at like an 8th grade level. So if that sounds interesting to you, stick with me and I'll see you on the other side of the music. Hey, gang. Welcome to the ChatGPT experiment. This is the podcast designed to help you better understand tools like ChatGPT so you can find a nugget of value for your personal or professional needs. My name is Kerry Weston and I'm your host. Hey, first time listeners. Welcome to the show. Welcome to the Community repeats. Welcome back. There is a lot changing, isn't there? There's a lot going on, there's a lot being said, a lot being done in AI and it's hard to keep up. And here's what I find. There's a lot of folks that use words and terms that we're thinking. We're supposed to know what they're talking about, but oftentimes we don't. And two of them, agents and vibe coding. I've read a lot about and I've heard a lot about, but you know, I was thinking the other day, I'm not quite sure I understand what these are. And the more people I talk to, I realize a lot of people feel the same way. They just kind of go along with it, thinking that they're supposed to know what these things are, but not really. So if you feel that way, you're normal. Now, there's some of you listening right now that are like, of course I know what they are. I use them all the time. Well, good. I appreciate that. And your curiosity is paying you well, but there's a lot that don't. Right. And so when I talk to curious beginners and people that are capable and curious, I think it's important to just have honest conversations about what we know, think right, and want to know about. And I think these two terms are part of that. And, and as I was reading some articles. I said, you know, I need to talk to somebody who can maybe explain this in a simple way. And so I reached out to my friend Greg and bring him back to the show. And Greg Howe is the founder of Gimme Info. We've known each other for decades and smart, smart guy. And I love Greg's ability of being able to get into the weeds, but to also back up and explain it in a non technical point of view. And so I had Greg come on and I said, listen, I just want to know a little bit more about agents and this thing called vibe coding. I'm hoping you can explain it so my audience could get some value, you know, even if it's just the ability of being, okay, I get it, you know, the ability to say, I understand what that is. So if someone brings it up, you know, or if you read it, you know, that's kind of thing. So not a technical thing today, not a how to. And go do these five things. This is really just a water cooler conversation to help you better understand what those two terms mean. Now you're going to hear later in the episode you're going to hear callbacks to some terms that we use all the time. Frameworks and interns and just talk to it and role playing. So I want you to understand that a lot of the things that I've talked about on this show, a lot of the foundational elements are valuable for advanced stuff. Okay, so look into the archives here, into the episode, I've got a role playing episode, I've got a just talk to it episode, I've got a framework episode. Right. I've got a lot of basics that we put together to do the advanced stuff well. And when I do trainings and when I have conversations, there's a desire for businesses to fast track the basics to get to. Can you teach us how to do this advanced thing? And when I dig in, I'm like, well, show me what you're doing on a day to day basis. And the fundamentals are missing, right? The frameworks are missing. And so it's really just I guess reinforcing the fact that true mindsets and understanding of a mindset and a framework will serve you well for the simplest of tasks and it'll serve you super well when you get into the more advanced stuff. And the only trick here is to think normally. It's not to think like a computer programmer, it's not to think like some advanced engineer. It's really slowing it down and thinking in smaller bits and bytes. So if you're confused about things like agents and vibe coding. You're normal. But know this, it's really all about breaking it down into simple terms. Talking, using frameworks, and using that mindset that we've been talking about all along. That will serve you very well. Okay, enough about me. Let's get to the conversation. I hope you're doing well. And as always, your curiosity is going to be the key to you having any benefit from ChatGPT or whatever tool you're going to use. Okay? So I'm glad you're here. And until we talk again, do stay curious. And here's my conversation with Greg Howell. Well, hey, Greg, welcome back to the show, man.

[A]

Hey, thanks, appreciate it. Thanks for having me on.

[B]

Yeah, I, you know, I was having a conversation, somebody was asking me about vibe coding. Two things in the same conversation. Agents. What the heck is an agent? Is what they asked me. And then they said, I'm reading a lot about vibe coding. Can you. Yeah, what is that? And I said, you know, I struggled. I gave him what I could, but I said, you know what? I got somebody. I'm going to reach back out to. Greg.

[A]

I know a guy.

[B]

Yeah, yeah, yeah, I know a guy. I got a guy. So just for those that haven't heard you before, Greg, quick introduction. Who are you, what are you doing these days and who are you doing it for?

[A]

Yeah, great. So, Greg Howe. I'm the founder of a company called Gimme Info llc and right now we're building AI applications for small to medium sized businesses around things that you need to take care of the pain points and whatnot. We don't go for unicorns, we go for practical solutions and trying to bring AI into the everyday. And we understand that, you know, like Carrie said, there's a lot of people out there that don't know what they don't know and they don't know the basics of it, but they know they need it. And so that's where I kind of come in and make applications.

[B]

Yeah, there's a, there's a. I think fire hose is the word I want to use. And we were talking about that earlier. There's just so much that couldn't be done that a lot of folks are in AI paralysis. You know, what do I start, what do I do? How do I trim it down, you know, how do I niche whatever, you know, so flip coin. Heads is vibe coding and tails is agents. All right, Agents it is. All right, so let's start with agents. The word agents has been creeping into a number of conversations. It seems to have different definitions depending on who you're talking to and what they're talking about. So could you give me, in your opinion, the 101 8th grade kind of definition of what the heck is an agent?

[A]

I'll do the best I can. So the quick 10,000 foot view of an agent is. And I'm going to get grief for this because there are purists that have a specific definition that I don't adhere to. But the idea of an agent is kind of what you would consider to be an AI internal. An agent has a job and you go say, here, here's some input. Go do your job and produce some output and let me know when you're done. And what happens is you can start to look at agents as employees. And so these agents can talk to each other, they can work together, they can aggregate. Sort of like a specialized AI tool that sits there and waits for conversation with you. And you say, hey, I need to go do this thing for my marketing, whatever it might be. And it says, hey, I know a guy. And the important thing to remember is that an agent's job is singular focus. It should be. So you have an agent that scrapes websites, you have an agent that generates LinkedIn posts. You have, they don't have to be one big prompt, right? Everything has one input, one job and one output. The output of one of those agents can be the input to another agent. So you can kind of chain these things together to get a job done.

[B]

There's a lot of people listening right now that are thinking, I still don't know, and where do these happen and how do you do it? So what's the difference between. Or listen, tell me. These are just synonyms, but what's the difference between using the word agent and using the word custom GPT?

[A]

Yeah, yeah, that's a great, that's a great point. I was kind of skirting around the edge of that. So a custom GPT, if you're not aware, is basically ChatGPT with specific instructions that you give it to get a job done, which could be like we just said, and that can be considered an agent.

[B]

What makes it an agent is that

[A]

it has an input, it has a job to do, and it has an output specific stuff.

[B]

Okay.

[A]

You know, if you've looked at the custom GPTs, they have a lot of things like I generate images, it doesn't do LinkedIn posts as well. You know, when you, when you look at chat GPT, you can ask it anything and it'll Come back with an answer on any topic. Yeah, These custom things are like agents that are really narrow, focused, specific job in hand. They have specific things that they do for you. So in a sense, those can be agents.

[B]

Okay, so what makes. Let me just ask the question so I can set you up. We've determined what makes it an agent, what doesn't make it an agent.

[A]

Okay. So agents also go beyond a task and begin to start using a form of intelligence to make decisions. So my job is to translate what Carrie wants into tasks that get the job done. So a job is a bigger thing that has steps to it. Each step would have an agent that does that step. Now, what happens is the AI can determine. I don't need to do step two because of some condition. I'm just going to go over that and go to step three. So there's this inherent intelligence. I hate using that word because it really opens a lot of scary doors for people. But it's not, it's literally just. It makes determinations based on conditions for your getting the job done. An agent kind of stands alone to get the job done. And I think it's closer to having a contractor come to your house where they, where you say, look, I need. Or a plumber, right? I need you to come here and fix these pipes. They're going to have a tool belt. You don't necessarily tell them which hammer to use, which wrench to use. They determine based on the conditions of this, of what it's dealing with. You simply say, I need those pipes fixed. And he or she makes that determination. So that's the same thing with an agent.

[B]

Do you see why this is confusing to so many people?

[A]

Absolutely. No, I, I totally get it. And that's the thing. It's hard to talk about it as agents. You want to look at them as pieces of AI functionality that have a task in hand. And what it could be is like for your custom GPT, you start there, you go type in something, and stuff happens. Maybe that custom GPT is step four of a bigger process that I have. And so my agent says, I'm doing step one, step two, step three, now I'm going to carry GPT. It becomes an agent in and of itself.

[B]

Okay, I know where to find a contractor. You know, I'll. But I don't know how to find an agent. So where do you find it?

[A]

That is a loaded question. Just like there's many definitions of agent, there's not a good answer for this question there. So here's, here's the nuts and bolts. Most agents you're going to create yourself because it has to do with a specific job you were trying to get done.

[B]

Okay.

[A]

The wide open market of agents doesn't exist yet. It's coming, but there's no one Toys R Us you can go to and say, hey, I need an agent that does this.

[B]

Where do I even start? Like, what do I do?

[A]

It's a great question. The answer I'm going to give is the same answer for both. I want you to really look at it from a business perspective and don't come at it from an AI perspective. You want to try to first of all solve a problem, don't look at it and go, I need this AI agent to do. No, if you start using AI agent as your solution, that's. You're missing. You're missing the point.

[B]

I get that.

[A]

Okay, so what you want to do is if an agent makes sense to solve the problem at hand, then you would have to go use. I would recommend you use whatever tool you like, whether it's Claude, Gemini, Chat, GPT and start saying, hey, I'm looking to create an agent that can do this task for me. That's step two of a bigger process. I'm new to this. Speak to me like I'm in 8th grade and help me build this thing out. It's going to give you current answers because there are many places and many ways that you can generate this agent. And what I really highly recommend everyone who's listening does is come at the tools and say, this is what I'm trying to solve and you do this for me.

[B]

Okay, so a couple things I just heard you say.

[A]

Yeah.

[B]

Is. And we had an episode on this a few weeks back, which is don't start with AI, start with the problem and work back on. Which I agree with. I'm happy you said that. If I use, let's say Claude or ChatGPT and I say I have a problem that I'm wondering if an agent would be right for, and maybe it's a paper process or whatever. It's a process on a computer and I'm new to this. I'm going to follow your lead here. I'm new to this. I really don't know what I'm doing. Speak to me like an 8th grader. Help me map out or help me think about how an agent could help me do this. Would I then go and just explain what I'm doing?

[A]

Okay, so now you branch into vibe coding, so.

[B]

Oh, well, hold on. That's the Other side of the coin. So is it time to move over? Is that what we're doing?

[A]

It is. It is.

[B]

Okay.

[A]

There's not much more that we could say about agents in specific. Vibe coding will play into that.

[B]

So agents. I can't define it, but I know it when I see it. It's basically what we're saying kind of a thing. It's. It's because it's nebulous at the moment, what you're saying. Because there's multiple things here. I get that. It's hard to define. It's a wild west. Any application.

[A]

Yeah. Look at it like a mini application. It's. It's a. It's a job doer will chat GPT

[B]

or Claude or Gemini. If we go down the path that you just shared, which was, hey, I'm thinking I need to solve a problem and I'm wondering if you can help me. I'm wondering if an agent, building an agent, finding an agent could help. I'm new to this. I don't know anything about it. I mean, I'm an eighth. You know, talk to me like I'm an eighth grader and then I kind of explain what I'm doing. Right. Yeah. At that point, having not done this before, is. Is the expectation that. Let's just say ChatGPT. ChatGPT is going to come back and tell me that I should go look at these other pieces of software or these other tools. Is that what you're thinking? Yep.

[A]

Absolutely. I would actually recommend that you start first from the concept of. Then you tell it this. Don't and say, say to it. Do not tell me a specific tool. Is this an AI agent? Is this something that I need to program? Is, you know, what you want to do is pretend that you're talking to consultant who, who can serve you the moon or can serve you a tiny little pebble. You decide and it's like, no, I need this problem solved. How would you solve it and make it easy? And the other thing you want to add to that is if with the expectation. And again, this goes back more to business process. But who's going to do the coding? Who's going to maintain it? And this is the thing about Vibe coding that I was going to talk to you about is there's a so much more. Once the agent is built, who maintains it? How do you maintain it? What. What about security? What about all these. All these bigger things suddenly show up. But you sort of have to know the direction you want to take as a business leader and Put that in at the beginning and say, hey, I need it to be this, this, this, and this. Where do I go from here? And step it through and don't let it tell you a tool, don't let it give you a final answer, and don't let it generate you code. Work out the concept so that you're like, I get it. That makes sense to me. Now how do I do that? And that's when you suddenly open the doors up and go, build something.

[B]

All right, so before we move on to vibe coding, let me just see if I can summarize what we just talked about and you tell me if I. If my takeaway is close. Yep. An agent is a lot like a contractor. It understands as a problem to be solved and it has maybe multiple ways of solving it, but it's using a purpose, it's purpose driven, it has an input, it executes and has an output. And it could be talking to one or many pieces of software or tools, but its ability to recognize what its purpose is for and apply some conditional logic to it through the. The way is, is the beauty of an agent. It's not a single hammer. A hammer is not an agent. A contractor that has a hammer but also has a saw and also has a wrench, in this kind of scenario would be considered an agent.

[A]

Bingo. Nailed it.

[B]

And the best way for us, from a simple beginner point of view, to understand if we even need an agent or if an agent could be helpful, is to open up the tool of our choice, like ChatGPT or Claude, and have a conversation with it to explain that you are looking into the possibility of an agent, but you don't know. You're looking for guidance. You want ChatGPT to talk to you in a simple way. And I don't want you to recommend software, I don't want you to recommend specific tools. I just want you to consult with me back and forth to see if what I'm dealing with, looking to solve the problem I'm looking to do here is even agent worthy. Is that, does that kind of sum up where we're at?

[A]

Absolutely. That's beautiful. Yes. That could be a checklist for people as they go into this. Because you want to come at it from the business perspective and you, you need to understand things. And the AI is going to be. What you just said is what I actually still do in my day to day. I use it in that exact same way to solve my own problems. The difference being I have experience, so I know the technical side, so I Don't shut that out right away. But I follow the exact same steps you just laid out.

[B]

Okay, so without telling me any details, technical or otherwise.

[A]

Yep.

[B]

Listen, two consistent themes here are, you know, start with a problem first and, you know, talk to it. So find your one thing. Right. If you can find your one thing. People are so overwhelmed with how much AI can do and how much AI is doing that they're forgetting just to go find that one thing to make their life easier. And we've talked about that multiple times, so I'm hearing the same thing.

[A]

You stack it well, because that's the thing, Kerry. Once you have an agent that does a good job, say, with web scraping, it solves the initial Legion problem, but it can also be used for other problems that you might want to solve. So these agents are reusable. You stack them differently.

[B]

That's all.

[A]

You put them together differently.

[B]

Yep. Yeah. My friend Chris Green, yesterday, on a call, said something that I thought was really simple and smart, which was, for every action, there's a reaction. And that's all we're talking about here, is there's an action needed, what's the reaction, and the reaction could become the next action. We just.

[A]

Exactly.

[B]

Okay, so let's get to the other one that comes up a lot. So hopefully somebody listening. You know, at the water cooler, if there's cocktail parties, whatever someone says, agents, they have an understanding to at least talk about the contractor mentality to agents and has enough to understand what we're talking about. I really don't know what Vibe coding is. I'll be honest with you. So I gotta ask you, let's go down the simple, similar path here to help somebody understand what's even being meant. When. Yeah, coders, every. If there's a programmer or coder listening, they're like, oh, this is too, too simple. But the rest of the world doesn't think that way, has no idea what we're talking about. So let's. Let's get into a simple definition.

[A]

Yeah, the simple definition is. So what it is is this. You describe what you want in plain English, AI writes the code, and then it becomes this iterative process where you say you want to change something and it changes it. And then you say, well, that's not quite right. If we move that over here, can we change this button? A lot of times, it's a visual program that you're building, can be an agent, but it's basically that whole concept of Vibe is I want it to do something like this, and it goes well, how is this it work? It generates code based on the sheer volume of code that it has in its background. Okay, so quick, quick. You're trying to just please.

[B]

So we. We've both been in programming environments in our past. Is this both for visual as well as functional?

[A]

Absolutely.

[B]

Wow.

[A]

Yeah. So the idea with this is that you're again, coming at it from solving a problem if you decide a program is what you need. What I do. And this is, I. I actually do Vibe coding myself. This. I did an entire SaaS product.

[B]

And what's that?

[A]

Sorry. Software as a service. So basically, people can log in and get their version of this software and branded for them. The whole. I never wrote a single line of code. The entire thing was Vibe code. And it works beautifully and it's better than I would have written it. It, like, actually is better. So.

[B]

So the reason I brought this up, the reason I even reached out to you in the first point, and I think I mentioned this in the first outreach, was Monday.com situation, which was the productivity project management tool, Monday.com the stock tanked on one particular day because a couple folks. I think it was a Wall Street Journal article that I was reading in about an hour, and maybe it was more than that, but the article said in about an hour, there were some folks that went down a Vibe coding rabbit hole and basically replicated the entire functionality of the software, Monday.com and that scares a lot of people. Right? Absolutely. Rely on revenue. So they just did the same thing. They just had a conversation about functionality and features and looks and all that kind of stuff. Is that what was going on there?

[A]

That's it.

[B]

I.

[A]

You know, because when you look at it, whether it's Monday.com or any of the other software that has the same job, you can articulate that job without talking about a single screen. You could theoretically throw a screen on there too. But it's basically. I'm trying to do this thing with a CRM. You know, it's like customer management, leads could be all. All kinds of things, but you can verbalize it. Once you can verbalize it, the AI can go. I've seen something that does that. And what it does, it. It's leaps and bounds from even just less than a year ago. Less than a year ago, I was working in a tool that I felt was kind of clunky visually. This one that I just created. Unbelievable amount of time savings and done. Looks like a professional designer did the work. The first. The first iteration was not. I said, this looks kind of old School. Can you make it more modern? 20 minutes later, I had a beautiful design and I was, I actually said out loud, wow. And clapped. I mean, it was that crazy. But that's where.

[B]

So we could combine these two things with what you started to say earlier is if I was a massage therapist on my own and I wanted, wanted to have an online kind of presence that took an application or a form, and that form went into something that sent out a video or sent out an email and maybe even booked an appointment. There are hundreds of software programs right now that I could go and rent, basically write my monthly subscriptions. That's a SaaS, right? That's software as a service.

[A]

Yep.

[B]

But if I had the patience and if I really wanted to, I could just have a conversation with a tool, chat, GPT or whatever you're talking about in a vibe coding mentality and create the same thing. Is that really kind of what you're saying?

[A]

Yes. And before we lose people that are going to go, well, I'm out of here, I'm going to go build my next thing. Let me just say that there are some caveats to it and they are huge ones that people don't like to talk about.

[B]

Let's go.

[A]

So in a general sense. Well, first, first of all, yes, absolutely, you're right. And I want you to notice that what you did was explain invert in words what you were looking for. That is exactly how you talk. Do the ums and the odds and the I think I want this and be. You don't have to be crystal clear. Give it the idea of where you're going. That's the whole vibe aspect is it's going to give you something and you can go, no, not like that. Make it more like this. You get a vibe, you get this whole thing going. This. And it takes, it does take iterations, but with the right focus, you can in an hour whip together some pretty impressive stuff. But what I want to say, the caveats are, don't suddenly, as a massage therapist, say, I want an entire CRM that has voice this and that and has a chatbot and a website and a this and a that. Don't, don't go there again, solve the first problem first. What's the painful thing that an application or an agent could solve for you. And then try vibe coding. Understand that what it generates can sometimes be crap. It shouldn't turn you away from it. It just say, you can say to it, that was crap. Do it. I, I want this to be more professional. I want it to look like Monday.com and it goes, oh, oh, let me try again. That's. It's an iterative process and you want to make sure that you keep iterating until you get it right. The downside of that is, as you've seen with regular text generation, so code is just text. As you see with text generation, when your conversations get particularly long, it forgets what you talked about. Way backpack.

[B]

The backpack is full.

[A]

Exactly. And it loses some of the thread and it loses some of the mission in the vision. So you want to make sure it's going to do the same thing with ku and so you want to make sure that you keep revisiting it. Some tools are better than others at this.

[B]

I don't want to get too far into the rabbit hole of how to do it, but I think that what I'm. What I'm. I love what you're just saying, which is maybe save some of it in pieces and trust and verify along the way to make sure that the newest version of the polish doesn't kick out the functionality that was so important in your first conversation. Something like that.

[A]

Yeah. I think it comes down to it's our expectations, Gary. It's like the nice thing is a lot of these tools now can connect to something called GitHub, which is where you put code so you can version these things. So you can say, yeah, go ahead, make a complete wholesale change and then go, oh, wait a minute, undo. And it will literally go back to the version before, which is really, really solid. But the thing is, I know that that is possible. The business person isn't going to go, oh shoot, undo necessarily. They might go fix it, you know, and move forward and make it worse. So these are caveats about some of this vibe coatings. Yep.

[B]

The backpack mentality, that's a, that's a big caveat. That's. Is there another you should think about or this is what you should know. Big topic along with the degradation of the memory.

[A]

So AI will confidently build you the wrong thing if you're unclear, it doesn't push back on you and go, are you sure? That doesn't really make sense. It's expecting you to direct. So again, the. One of the best things you can do with that to offset that is to a know what you're asking for, like in a general sense, in a visual sense. Like if you like Monday.com, say I want it to be like Monday.com in this area. And it will, it will do its best to meet that. That's going to be very helpful to you. Give it examples, give it things that work. You want to be clear up front about what you're trying to accomplish, but not necessarily.

[B]

And you might not know, you might not even know until you see it. And that's just right.

[A]

I mean, but what I mean is about what to accomplish from a business sense. Like, don't tell it I want a website. Unless you know, you want a website. Like I, when I start my own applications, I say to it, I'm trying to do this for a customer.

[B]

Yeah.

[A]

And I let it tell me a website makes a lot of sense here. Or an agent makes a lot of sense here. Like the SaaS product that I'm building was simply because the AI said this sounds to me like a SaaS. My head wasn't building a SaaS. My, in my head I was building a client specific, put it on their system kind of application. And it said, no, that's going to become a maintenance nightmare down the road.

[B]

So, so there's an opportunity here to tie in some basic principles from ChatGPT101 from the very first days of these, of this podcast, which is, you know, talk to it, talk to Chat GPT and tell it what you need. Give it a role to play. Maybe it plays an advisor role or in a consultant role and it's going to interview role play a little bit. So have it ask you questions, have you go through the interviewing process. Maybe you are a developer. You could say, play the role of developer. Tell me what I don't know. Ask me questions based on what needs to happen, think about things I couldn't think about. So all of the practices that we've talked about, just talking to it, tell me more. Role playing, give it a purpose. That all kind of comes into play here, doesn't it?

[A]

Absolutely. Because it's no different than when the early days and you'd say, hey, write me a poem about pizza. It would write you a poem about pizza. You can do the same thing with Vibe coding. Hey, build me an application that's a CRM. It will build one for you.

[B]

Yep. It's like asking the right questions or having it ask you the right questions or at least understand the context and purpose and small steps. That's going to make a major difference.

[A]

Yeah, absolutely. And the whole Vibe coding concept, the thing I want people to be aware of with Vibe coding is it is still coding and you are going to save a lot of money and time on developers and designers and the I got to spend $30,000 with an agency just to find out that this prototype doesn't actually do a good job. It doesn't do what I would expect. Like these are, you want to vibe code internal things and prototypes and get it to a point where it works. But then you really do want to pull in experts at the end because there are a lot of things about development that you rightfully don't know. The same way the developers don't know how to run your business the right way, how to do the logistics of your shipping, right? It's like you know this stuff, but you, you are going to be having this code generated for you and there's a lot of things that you don't know that you don't know. So one thing that comes to mind

[B]

right now, if I'm using all of these things together and we'll just go outside of software in a practical sense. If you were designing a house, right, typically you'd start with an architect or designer and they would ask you the first question, right? Like how many bit you know, they'd start with the foundation and work up, right? How big are we talking? How many bedrooms, whatnot? It almost sounds like you could, in this case you could just say, hey, chat GPT, I want to build the house. I want you to be the architect. I want you to be the designer asking the questions I don't know. And you could end up with. And I'm not talking vibe coding, but I'm just putting this into one on one type 8th grade mentality. You could end up with a set of blueprints and architectural plans, right. That you wouldn't want to give to a contractor and say, go, I wouldn't trust that. But you could probably take that and give it to a contractor or an architect and there might, that might be 90% good, right? But they could start at that 90 level. So you're so much further down the road than you would be if you started with who are you and how much is it going to cost to build a house?

[A]

Right, Yeah, I mean, you want to tie in the other part. Create an agent that is your architect. Create an agent that is your senior developer create. And that doesn't build code for you, but tells you here are the gaps. Well, I hear what you're telling me, but here's the things that you haven't mentioned. What are you going to do about this? What are you going to do about that? So you can have these agents or Even just custom GPTs, or even just chat GPT alone where you say to it, hey, I'm trying to get this accomplished. Tell me what I'm missing. Yep, highly recommend that Avenue. And then you could, in theory, say, provide me with an output that I can put into Claude code and let it generate this application for me and

[B]

give me a 3D rendering of the house that we just built.

[A]

Yeah, exactly.

[B]

So I figured out that I put the. I put the toilet on the ceiling. I didn't mean to do that with that.

[A]

Well, that's a tough one, especially when you flush. But with the thing like with this. That's exactly what I did, is I said, before you write any code, I want to see what this page looks like. What does the main page look like?

[B]

Yeah.

[A]

And I was like, not only did it. It didn't just generate a PNG graphic for me, it actually generated an HTML mockup that I could mouse around. Like, I just double clicked it in the file system and it pulled up a webpage that had. It was graphics in there. It didn't work, but it allowed me to mouse over things and click on things and it changed pages. So it gave me a working shell and goes like that. And I was like, yeah, dang it. Just like that. And so then it can fill and fill it all out.

[B]

So we started off with two big questions. What is an agent? And what is Vibe coding? We wanted to make sure that people understood enough to have a conversation. And when they read it and see it and hear people talking about, they could know the agent is the contractor. And the Vibe coding is the way in which you can talk in plain English to get something really technical done that might have features, functionality, and a little bit of polish. Then you really don't have to be in the software anymore. Design and coding, you use English or whatever language you have to kind of talk to a tool. And that tool recognizes how to do the detailed programming work. That's. And it's all by feel. Like, I think I like, I want more of this. I didn't want any of that. That's where the Vibe part of that comes in. So.

[A]

Absolutely. That's really. Yeah, no, you just nailed it. I like the way you summarize both of those. That's it exactly.

[B]

I love it. Well, Greg, if people want to find out more about what you're doing or even reach out, what's the best way for them to do that?

[A]

The best way is probably to connect with me on LinkedIn and you can go to LinkedIn slash, gimme info, or you can just look up Greg Howe and you'll see me out there for with Gimme. And that's probably the best way to contact me. If you go to my website right now, it will show my estate IQ SaaS that I'm building. It's really my focus right now. Instead of being a company of, hey, I build all kinds of things AI, I'm like really focusing on this one product and then expanding from there. So. But that's the best way to get in touch with me. You can always email me Greg, at gimme g I m e e.info so remember, it has two e's on the end of Gimme.

[B]

I'll put that in the show notes as well.

[A]

Yeah, yeah, that's always a good way to.

[B]

Are you vibe coding any of this estate stuff?

[A]

All of it.

[B]

Wow.

[A]

The. The entire SaaS, I did not write a single line of code. I spent probably 10 hours massaging it, changing it, looking at it, going, this doesn't quite work the same way. I had to do some really convoluted things and then said, no, scrap all that, go back to version one. And it did just fine. And I'd be happy to share some of the ups and downs of that process with anybody who wants to go one level in and start getting their hands dirty. You'd be happy to share what I've learned through that. It is a process and that's the thing I really want people to understand. Get your hands dirty, but understand that it's just like the contractor mentality when you start messing with your house. There are certain things you think you can do that really you should probably get a professional help. But there's a lot that you can do that you didn't realize you could.

[B]

Yep, that's really cool. Well, thanks. I know there's a lot of curious folks here that are probably going to dabble a little bit, but for the most part, having folks be just informed enough to talk about it and understand it is a big benefit. And I appreciate you sharing. You're going to help a lot of people in that regard. So, Greg, thanks for being here. I appreciate it very much and we'll to you soon.

[A]

Yeah, thanks, Gary. Thanks for having me on.