Episode Categories:
Resources:
Join the discussion on LinkedIn: Got something to say? Let us know on LinkedIn and network with other AI leaders
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
Start Here Series in our Inner Circle Community: Join for free access
The Vibe Coding Boom: How AI-Generated Applications Are Reshaping Enterprise Software—And What Business Leaders Must Know
A seismic shift is underway in enterprise software development. The rise of “vibe coding”—the practice of building applications by describing requirements in plain language to AI agents—has dropped the technical and financial barriers to application creation. Yet, while the early promise includes dramatic speed and cost advantages, the approach brings its own risks, from hidden technical debt to costly rescue engineering. These are not passing trends—they’re structural changes already forcing billions in IT reinvestment and redefining talent needs at the executive table.
Vibe Coding Platforms: AI-Powered Application Builders Defining the Market
The cost and time to create new software has collapsed, with estimates that 41-46% of new code is now AI-generated. 92% of US developers actively use AI coding tools at work. Key platforms like Claude Code, OpenAI Codex, Cursor, and Google AI Studio have made it possible to generate entire codebases, deploy working apps, and even automate product research—all without traditional software expertise.
The spectrum of vibe coding platforms covers:
Terminal or Command Line Agents: Tools like Claude Code and Codex operate directly in a developer’s terminal, reading codebases and autonomously rewriting code at scale. These solutions provide multi-model support, parallel agent execution, and the ability to conduct in-depth analysis or codebase overhauls.
AI-Native Code Editors: Solutions such as Cursor, GitHub Copilot, and Windsurf embed AI directly into the integrated development environment (IDE), supporting background agents and multi-user, collaborative editing. Cursor’s multi-model approach lets users select from state-of-the-art models like GPT-5 or Gemini.
Browser-Based App Builders: Platforms like Lovable, bolt.new, and Google AI Studio are optimized for non-developers, converting plain language prompts into fully deployed SaaS applications, complete with authentication, payments, and built-in security scanning.
Vibe Coding Business Impact: From Prototyping to Core Systems Replacement
Enterprises are bypassing traditional development cycles. Teams that once measured product prototype timelines in weeks now create functional prototypes in hours. In documented cases, companies have replaced commercial CRM systems (such as migrating from a $40,000 Salesforce contract to an internally generated CRM built with Lovable in three hours) or slashed product prototyping from six weeks to mere hours.
This efficiency enables business lines to launch new services and revenue streams without waiting for overburdened software engineers. For organizations with previously limited development capacity, it opens access to modern digital product launches and customization on demand.
Vibe Coding Security & Technical Debt: Hidden Risks for Enterprises
Beneath the speed are structural risks. The majority of vibe coders—63%—have never written code manually, resulting in software built without foundational understanding of architectural principles, security protocols, or compliance requirements. This has led to production incidents ranging from single-misconfiguration data leaks (such as an AI-built agentic platform exposing over 1.5 million API tokens due to a database error) to system collapse when apps hit real usage.
Research shows trust in AI-generated code is volatile; developer confidence dropped from 43% to 29% in one recent cycle, with top frustrations centered on flawed code that “looks almost right” but fails edge cases or security standards. AppSec studies reveal that 25% of sampled AI-generated codebases contain confirmed security flaws, emphasizing the need for new review and governance models.
The 90-Day Black Box: The Need for Rescue Engineering
The business cost of unchecked vibe coding surfaces 60 to 90 days after deployment—when neither AI nor original non-technical builders recall application logic or context. Rapid-fire delivery gives way to “black box” architectures, halting further iteration or requiring expensive rebuilds. Thousands of startups now face $50,000 to $2,000,000+ refactoring projects to replace vibe-coded MVPs held together by what seasoned teams describe as “duct tape, hopes, and dreams.”
As a result, "rescue engineering"—the post-deployment task of auditing, restructuring, and securing AI-built systems—has emerged as a high-growth specialty, increasingly essential as legacy technical skills no longer match the AI-driven pace and complexity of new application stacks.
Vibe Coding Governance: Strategies for Sustainable Advantage
Market consolidation is creating clear best practices for sustainable AI-driven development:
Governance and Oversight: Treating AI-generated code with the same process rigor, review, and documentation as hand-written applications.
Built-In Security: Selecting platforms that include penetration testing and security review modules (as pioneered by recent Lovable updates).
Context Management: Prioritizing platforms with high-context windows and version tracking, ensuring that both code and business logic remain accessible for audit and update.
Hybrid Human/AI Teams: Structuring teams to combine technical oversight (even where non-developers initiate builds) with prompt access to "rescue engineering" resources for ongoing maintainability.
Outlook: Vibe Coding as Permanent, Not Optional
No business can assume that vibe coding is a fleeting trend. The movement is evolving into “vibe working,” with platforms like Microsoft Copilot Cowork and Google AI Studio democratizing software creation for every knowledge worker. Leading firms are not just using these tools—they’re overhauling their governance and talent strategies to mitigate risks of technical debt and seize a lasting cost and speed advantage.
Executives seeking to capture value will invest not only in AI platforms but also in developing procedures around code review, security, and lifecycle maintenance—curbing the burst of productivity before it compounds into future liabilities. The winners will treat vibe coding as a core business process, not simply a technical shortcut.
Topics Covered in This Episode:
- Vibe Coding Definition and Industry Impact
- Top 10 Vibe Coding Platforms Overview
- Terminal and Command Line Coding Agents
- AI Native Code Editors Breakdown
- Browser-Based App Builders Comparison
- Vibe Coding Adoption by Non-Developers
- Security Risks in Vibe Coding Applications
- Rescue Engineering and Codebase Rebuilds
- Developer Trust Collapse in AI Coding
- Strategies for Sustainable Vibe Coding
Episode Transcript
Jordan Wilson [00:00:16]:
Imagine that you can describe your dream house out loud and poof, a building appears in front of your eyes in minutes. It looks incredible. It's exactly what you asked for. And you move in. It's awesome. You tell everyone how easy it was, but then almost instantly you wanna make some changes. You wanna add a second floor, heck, maybe even an outlet here or there, or just maybe you wanna hang a towel rack, but then you run into some problems instantly. You realize that no one knows which walls are load bearing or if the electricity works, and turns out that those floors look sturdy.
Jordan Wilson [00:00:54]:
But when you plopped your couch down on those marble looking floors, turned out that they were kind of paper thin and they started to sink. And it turns out the whole structure was great in theory, but it was held together with duct tape, hopes, and dreams, and there definitely wasn't a permit. And now you're in trouble. And this scenario is exactly what's happening right now across the enterprise in the software industry at a scale that we haven't seen before because of vibe coding. But is all vibe coding bad? Is it all smoke and mirrors and hallucinated code that's held together by strong models and agentic loops, or can you actually Vibe code a production ready app without much coding knowledge? Today, we're gonna be giving you those answers and breaking down the Vibe coding boom, the opportunity, the potential crisis, and why none of this is going away. Alright. Let's jump into it. Here is the big picture of what we're gonna be covering on today's show.
Jordan Wilson [00:01:53]:
Because you can't escape the reality right now that AI has completely collapsed the cost of building software, but it's not all good. Right now, it's estimated that about 41 to 46% of all new code written globally is now generated by AI. And right now, the latest update that we've seen, which is actually a pretty old one, is that 92% of US developers use AI coding tools at work every single day. But even Andre Acarpathy, who coined the term vibe coding, has now declared it passe after only one year. It's because we realized that it's not always all good. So on today's show, here's what we're gonna be diving into and what you'll learn if you stick with us for the next twenty ish minutes. You'll learn the top 10 Vibe coding platforms that you should be paying attention to and their main benefits. You're gonna learn what happens inside AI built applications at the ninety day mark that triggers a billion dollar rebuilds, and you're gonna know the specific strategies that separate Vibe Coding winners from the casualties.
Jordan Wilson [00:02:58]:
Let's get into it. Welcome to Everyday AI and our start here series. Yeah. After 700 plus episodes, I legit had no answer when new listeners would come in and be like, Jordan, you have a lot of podcasts. Where do I start? Well, you start here, preferably in order, but the start here series is the essential podcast series to both learn the AI basics and to double down on your AI knowledge. So if that's what you're trying to do, make sure to go to starthereseries.com. That's going to give you exclusive free access to our inner circle community. It's the only way that you can get in right now, and that's gonna pop you right in the middle of our start here series, space there, and you can go back, read, watch, and listen every single, episode in order all the way from volume one to today's episode, volume 18.
Jordan Wilson [00:03:45]:
And if you missed our last episode, like I said, it's best if you listen to them all in order. In volume 17, we talked about the responsible AI playbook and what it means and five moves to ensure your AI strategy survives. And that leads us straight to today because if I'm being honest, one of the things that when you talk about, AI strategies, so many enterprises' new AI strategy is essentially vibe coding. Right? Whether it's creating software for themselves or for, their consumers or just vibe coding completely new ways to work internally. It's pretty easy. So if you are like, wait, can you hit rewind here, Jordan? I'm brand new. I got you. Like, vibe coding is this.
Jordan Wilson [00:04:31]:
You describe something in plain English or whatever language that you speak, and an AI builds software. Right? So it handles all the database, the deployment, the coding, even certain platforms handle the authentication, payment platforms, all of that. Right? So this is a term, that OpenAI cofounder Andrej Karpathy coined in February 2025, which is crazy to think about because it seems like we've been talking about vibe coding for, like, five years, but it's been, like, fifteen months. Right? And I kind of called this out a month before, you know, kind of the vibe coding, terminology even existed in our, twenty twenty five AI predictions and roadmap series. And I talked about at the time that nontechnical people would be building themselves software by the year end, and I think that definitely came to fruition. But now fast forward into 2026 in the Vibe coding platform has exploded in the scene. And there's a handful of even different types or categories of Vibe coding. And like I said, this is an exploding sector.
Jordan Wilson [00:05:36]:
So, you know, as an example, Cursor, right there, an AI powered code editor, and they've already hit $2,000,000,000 annual recurring revenue in just a few months after, you know, their last earnings report. But most of the people fueling all of this growth are not even software engineers as well, or at all. So yes, you can have traditional software engineers or developers who are Vibe coding, and you've seen this as well. You know, some, interesting but kind of, I don't know if heart wrenching is the right word. Right? But talking you know, reading these articles about, you know, software engineers with twenty plus years of experience, and they're like, wait. These AI platforms do all of this now at a much higher level better than I ever can. Right? So it's almost got to the point now where, you know, if people know what they're doing, a lot of times, experienced software developers aren't even writing code anymore. Right? If you look at Anthropic, which you could say is potentially the leading, software dev agency in the world, right, because their, products called code, Claude Cowork are some of the best, out there.
Jordan Wilson [00:06:46]:
Right? They said they don't even write code anymore themselves. So, first, before we get into kind of the good, the bad, the ugly, and why I don't think vibe coding has go away, let's very quickly now that you have a basic definition of vibe coding, I wanna talk about some of the, 10 maybe of the most popular, kind of vibe coding tools. And they are in different kind of categories. And these categories, well, I define them. There's no, exact definitions that the industry, adheres to. And there's definitely a lot of crossover and, you know, you can make a lot of arguments, but I'm trying to simplify it for, our audience that's beginning. So first, I'll say that we have the terminal or command line coding agents. So if you don't know what a terminal is, computers have a terminal, and you can control your computer through the terminal.
Jordan Wilson [00:07:32]:
Right? And it's becoming more and more impactful now as we see, you know, platforms very popular like Claude CoWork that you can literally run things on your terminal. Right? So as an example, I always tell people, oh, I always have, you know, codecs and quad co work running, and people are like, tell me what you have doing. Right? So as an example, I have quad code going right now. It's going through my 750 plus transcripts, and I'm thinking about starting a new kind of episode series, where I piece together multiple interviews that I've done. I've interviewed hundreds of people over the years, and I've always thought, man, it'd be cool to kind of combine, you know, have three different people, older interviews, splice them together. You know, I can go in there and, you know, narrate between. So I have Claude CoWork as an example going through all of those local files on my computer right now, doing some research for me at the same time and putting together some ideas that I'm gonna go through and, you know, work with. So that's an example of a Vibe coding platform.
Jordan Wilson [00:08:31]:
Right? Claude code, Claude co work using the terminal. So runs in your terminal. It can read your whole code base and make autonomous multival changes. And this is also important to talk about before we get into these categories and the 10 most popular apps is it's not just for coders. It's not just for software development. Right? And I think that we'll start to see this probably in the second half, or the very 2026. And I think it's not really gonna become apparent until we get the new super app, from ChattGPT where they're essentially gonna be, combining ChatGPT, their browser, and codex. Because I've been saying codex is not really well, it is a coding platform, but it's so much more than that.
Jordan Wilson [00:09:11]:
So keep this in mind as we go through these. Even if you're not a software developer, a coder, right, I think it's important to pay attention not just to the rest of this episode, but to these 10 different, kind of platforms as well. So first, Claude code. I'd say it's probably the most popular and one of the most widely used, kind of vibe coding platforms. Although, yes, it is much more than traditional vibe coding. So there's a lot of great new features that have been added. And, again, depending on when you're listening to this, you know, so if you're listening to this in December 2026, you know, some of these new features aren't gonna be very new. Right? But if you're listening in April, right, things like computer use, auto mode, Claude or sorry, cloud, auto fix for PRs, a lot of new things, that have been, released in Claude Code.
Jordan Wilson [00:09:58]:
And I think this is kind of the terminal agent that just reads and rewrites your entire code base. Right. The other thing that I've liked recently is the new 1,000,000 context window by default in Claude code. So, by far, it is broadly considered the current leader in agentic coding quality and code based reasoning, but our second on the list here is catching up. So OpenAI's codex, you know, they recently announced that they have 2,000,000 weekly active users in codex, which is pretty impressive for a product that's only been out for, like, less than two months, at least the desktop version. So this is kinda your multi service, multi service ancient system across the app, command line, the IDE, and the web. So you can use, you know, codecs in a lot of different places. And, kind of what's new or what's worth it is, well, the model, g b d five four.
Jordan Wilson [00:10:47]:
I'm always in codecs using extra high, reasoning. Yes. It takes a little bit longer, but in my opinion, right, if I'm running the exact same task in quad code, versus codex, quad code is gonna be way faster, and it's gonna look way better. But usually, codex is gonna have way fewer issues. Right? And that's the thing too with these coding, vibe coding platforms. You can work, a lot of them in tandem. Right? When I'm working, I usually have Claude Code and codecs working on the same thing. Not at the same time, you know, I may have one with kind of read only audit access, passing things over, to the builder agents.
Jordan Wilson [00:11:28]:
But, codecs, like I said, the desktop app just launched in February. And one of the cool things that they've announced recently is, the ability to run multiple sub agents in parallel. Alright. There's a lot of other that might fit in that category of the terminal or command line interface coding agents, but I'd say those are the two big ones. Now let's move on to the AI native code editors. So these are essentially full editors that are, rebuilt around your AI. So these are for developers probably a little bit more than even Claude Code and Codex who want AI inside of their existing workflow. So the big one here probably is Cursor.
Jordan Wilson [00:12:05]:
This is kind of an AI first v s fork code with background agents in multi model support. That's the big thing right there. Right? Multi model support. Whereas, obviously, you know, codex, by OpenAI and Cloud Code by Anthropic run those respective models. So, like I mentioned, Cursor did hit 2,000,000,000 ARR in February with over 1,000,000 paying customers, and then you can use the best models. Right? So you can use OpenAI's GPT five four. You can use CloudOpus four six, Gemini three one, or Cursor's own composer model. Next, Wind Surf.
Jordan Wilson [00:12:41]:
So Wind Surf went through kind of a a weird 2025. There's a bunch of these rumors. Right? Oh, who's gonna, you know, who's gonna, acquire them? You know, OpenAI was gonna acquire them, but then they didn't. But then Google essentially aqua hired, their CEO and some of their cofounders, and then Cognition acquired the rest of the company. But, it is a great agentic IDE. Right? And right now, that wild, 2,400,000,000.0 acquisition saga. And sorry, I I maybe skipped over a couple of definitions. I don't wanna, leave you all here with alphabet soup, especially in our start here series.
Jordan Wilson [00:13:19]:
But what I'm talking about, you know, I already defined vibe coding, you know, command line interface CLI. Right? So that's essentially interacting, with a Vibe coding tool, through a terminal, kind of a a different way to use the terminal. Now when I say an IDE, that's an integrated development environment. So that's essentially like a toolbox for writing computer software. Right. So it's a single application that combines all the essential tools that a programmer might need, into one place. So they don't have to switch between a bunch of different programs to get their work done. Right? So to simplify it, that's what an IDE is.
Jordan Wilson [00:13:53]:
Right. So, Windsurf is definitely one of the more popular ones, but probably the most widely used one, even more so than curve cursor because of its parents is GitHub Copilot. So that's Microsoft's, you know, kind of AI powered Copilot, coding tool. And this definitely is the broadest IDE support at the lowest premium price point because, it's only $10 a month, and they also have free tiers that give you 2,000 completions across all major IDEs. And this kind of functions more as in line code completion versus the full, kind of vibe coding experience. And then you do have the, option, to have the enterprise version. However, we can't even talk, about vibe coding. Well, number one, just because GitHub, right, GitHub is kind of like the I say it's like the online Dropbox.
Jordan Wilson [00:14:47]:
Right? How, a long time ago, like, Dropbox was kind of the the big category for online storage. So GitHub is kinda like the online storage for, well, a lot of coding, but by coding platforms as well. And then Microsoft's GitHub Copilot, right, kind of their, AI tool of choice there. And then last but not least, the biggest category here, might be browser based app builders. And these are probably your more traditional fully vibe coding platforms. Although the first two categories, you know, you can by code anything, right, from zero all the way to app in those other platforms. But I'd say the browser based app builders are maybe a little bit more popular. This is kinda your prompt to app platforms.
Jordan Wilson [00:15:25]:
You describe what you want in English, and you get a working deployed application. So, I'd say probably in this category, lovable might be one of the more better knowns, and this is essentially prompt to deployed full stack web app for non developers. They have a new plan mode, which I think has been pretty popular that previews what the AI will build before writing any code. Right. A lot of the other platforms have that quad code has it, codecs has it, etcetera, but that's really helpful, especially, you know, if you are using, or paying for these platforms via an API. Right? If you're plugging in your own, you know, especially these multimodal, or or sorry, multimodal, vibe coding tools, if you're using your API keys, right, some of them, you just get a, you know, a monthly plan in different models that are more powerful, use more credits, but sometimes you're just plugging in your API key so they can get expensive. So it is helpful to have something, like the plan mode. Lovable also added, recently built in penetration testing, which is super important, and we're gonna get to that, later on.
Jordan Wilson [00:16:26]:
And that was kind of a first among Vibe coding tools. Next, bolt dot new. I'd say this is probably one of the fastest kind of paths from prompt to working prototype in your browser. And I also know there's so many other Vibe coding tools out there. A lot of them even have become more popular late twenty twenty five, early twenty twenty six. It was kinda hard once I got to, like, number eight. You know, I probably had, like, six of them that could have been, you know, number eight, nine, or 10. So a couple good ones were left off the list.
Jordan Wilson [00:16:55]:
Right. But bolt.new, definitely was probably one of the more popular ones in, 2024. So this runs entirely in your browser using web containers technology with nothing to install locally. They also have a great fee, free tier plan, and I think this is best for speed demos and quick MVPs. Although a lot of users say that overall, it's maybe weaker code, code quality than some of their competitors. Big name in this space, obviously, Replit. So that's their, you know, cloud IDE combining an editor database hosting and AI agents. So they just launched, the agent launch with 10 times more autonomy, building apps across 50 plus programming languages.
Jordan Wilson [00:17:34]:
I'm a big fan of the, the Replit agent. They also have effort based pricing that scales with actual compute and real time multiplayer editing makes it probably one of the stronger platform for learning and collaboration. Next v zero from Vercel, probably one of the cleanest React and Next. Js code from any AI builder. So this generates production ready components, you you know, using Tailwind CSS by default. And this is kind of evolved more into a full stack JS builder over time. And this is maybe the least lock in of any platform because you can export standard React code that then you can fully own and use. Alright.
Jordan Wilson [00:18:13]:
And then last but definitely not least, one of my favorites that I use probably the most, right, between codecs, quad code, and then Google's AI studio. Right? I I I I do use, Google's anti gravity. That's another one that didn't make the list, but I probably use Google's AI Studio a little bit more than anti gravity just because I've been using it for longer and a little more familiar. And, in March, Google did add anti gravity agent supports, inside, Google AI Studio, Firebase integration, and multiplayer support in there as well. And the cool thing and the simplest way to explain this, well, it's the free. Right? They have a great free plan, probably a better free plan than any of the others that I mentioned. It's a free full stack vibe coding platform powered by everything Gemini. So you can literally, in, like, five minutes, you can create a, you know, your own version of something like Neto Banana.
Jordan Wilson [00:19:07]:
Right? And then make it, you you know, specialize for your niche. Right? But, obviously, if you're, in your app, if you're using any of Google's AI, you do have to, you know, provide an API key to power that. Right? It's not like you can just use nano banana for free, but you can literally create your own version of any of Google's, you you know, AI. And like I said, they do have a generous free tier, and the the the team over there at Google AI Studio, probably all of them aside from maybe, Claude Code, and codex kind of, you know, though those two and then Google AI Studio are just getting so many updates so quickly. Alright. So now you know the big tools. You just need to know that this is a growing industry, and it is not going anywhere. If you thought Vibe Coding, was gonna come and go, and it was just gonna be some trend.
Jordan Wilson [00:19:56]:
No. Because Vibe Coding is turning into Vibe working, and that's where everyone's going. I mean, even just look at Microsoft. Right? Look at some of the things that they've, introduced recently, like, their Microsoft, Claude or sorry. Their Microsoft Copilot Cowork, which is powered by Claude Cowork. Right? So it's essentially a vibe coding platform. You know, Microsoft actually has a couple of others, but, you know, it's coming to the common knowledge worker. Right? And these platforms are growing quickly.
Jordan Wilson [00:20:28]:
So but there's there's an issue because right now, 60 a a study shows that 63% of Vibe coders have never even written a line of code, right, themselves, which is it makes sense. Right? Because, yes, you do have experienced developers and, you know, people maybe with a a a little bit of a crossover background, in software development using these. Right? And maybe it's allowing them to work in a language that they've never written before. But, you know, it kind of started with, okay, people, you know, using this. Maybe they're, you know, more commonly web developers, and then they're using it to, you know, launch React apps or something like that. But, eventually, it's turned into, well, you have people who don't know what code is using these. And these are builders that are just making quick prototypes. They're literally trying to replace enterprise software entirely.
Jordan Wilson [00:21:20]:
And it's actually happened. Right? So you have all these stories of, like, okay. Yeah. You can go vibe code something and and use it. That's cute. Right? But you're never gonna, you know, replace an enterprise, piece of software. Well, companies have. Right? I mean, we'll check-in with them and see if that's successful, but, you you know, one, kind of famous example is, hopefully, I get this company right.
Jordan Wilson [00:21:42]:
Atanam, their head of finance built a full CRM on Lovable reportedly in three hours that they used to replace their $40,000 Salesforce contract. Right? Zendesk's, product team cut prototype timelines from six weeks to three hours. Right? So, obviously, you have a lot of, vibe coding platforms where product teams or software development teams, have gone from, you know, timelines that were multiple months to, well, less than a day because of Vibe Coding. So I do think it's a it's a very powerful tool when used by people who absolutely know what they're doing. But just as the market is is reaching this everyone's vibe coding, developer trust in AI code is just collapsing. Right? And that's why there's a huge downside to just vibes. Right? And going back to that, you you know, my little analogy at the beginning of the show, right, if you could just talk your dream home into existence, but you walk in and, oh my gosh, the floors are sagging. That's the risk that you run because it's not all good.
Jordan Wilson [00:22:54]:
Yes. If you know what you're doing, vibe coding, obviously, is a not just a game changer in terms of what companies can produce. Right? Companies that already had software development teams, but companies that didn't. Right? Now all of a sudden, they can offer new products, new services, new revenue streams to their current clients and customers. Right? But it doesn't always work out well because if you don't know what you're doing and you still push something to production, it's obviously extremely problematic. And that's one of the reasons why trust has gone well down. So a Stack Overflow study show that confidence in AI generated code actually dropped from 43% to 29% in February. And one of the top frustrations for developers was that AI solutions are almost right, but not quite.
Jordan Wilson [00:23:45]:
Right? It's almost like, you know, there's more than, you know, one way to write a specific piece of code. Right? There's more than one way to design a certain website. And sometimes what we've seen and maybe this is, with some of the companies trying to get more token efficient and not necessarily get things right again. That's why, at least for me, sometimes I personally prefer, you you know, using, you know, codecs, g p d five four extra high. Yes. It's way slower, but it seemingly gets more things right with fewer issues. Right? But that's one of the biggest, problems right now is sometimes it might seem, like an AI coding app will, you know, the floor. Hey.
Jordan Wilson [00:24:28]:
Visually, it looks right if we make a cardboard floor or paper thin floor, but it's not gonna function correctly. Right. And I think the industry realized that it was hopelessly hooked on a tool set that it did not actually trust because AI doesn't really care too much, at least by default, about security. That has changed a little bit in 2026, but 2025 was a straight up security nightmare. Right, from, AppSec Santa, 2026 study showed that out of 534 AI code samples, about 25% had confirmed security flaws. Right? That's a very high percentage. So if you don't know, as an example, if you don't know how commercial plumbing works in your, you know, vied coated house, you're not gonna know that those AI built pipes were paper. Right? Good example of this, which is Molt Book.
Jordan Wilson [00:25:24]:
Right? So that was an AI built social network for AI agents. Well, accidentally exposed 1,500,000 API tokens from one database misconfiguration. So this is just examples of what can happen if well, especially as we talk about, agentic coding and, agents coding in loop with sub agents autonomously. Right? It's a little different when you have, an experienced human, kind of launching the code and testing this out. But when we go straight up agentic, I think that's where we start to run into issues. And when it starts to get, you you know, a little bit different, you know, kind of the, prospect of demos over memos, I don't think anyone will necessarily complain about. And I do think that's where we're headed. And I think that's why if you're even a nontechnical builder, I think in the future, it's like no one's gonna want a deck.
Jordan Wilson [00:26:13]:
No one's gonna want an email. They're like, just just build me a a demo of this now. Right? Because you can do this. Right? I've actually done this before, on calls. Right? Someone will be talking about something and be like, here. I just built this. Is this what you're talking about? But if you don't know what you're doing, maybe it's just a a demo. Right? Maybe it's not something for production, because that's where you will run into issues.
Jordan Wilson [00:26:37]:
And another thing is, well, it just generates code so fast. Right? Newer models. Once you get over the thinking, the chain of thought, and once it starts spitting out code, it goes so quick. So, studies show that fewer and fewer experienced developers are actually reading the code that AI outputs. Right? Some of them, the more famous, you you know, kind of, developers of the recent of of of the last couple, you know, months or quarters have said that they don't even review the code. You know, it was, Peter Steinberger. So the OpenClaw creator, who was hired on by OpenAI. So yeah, if you don't know OpenClaw, it is the most popular, piece of open software, open source software literally ever.
Jordan Wilson [00:27:32]:
And he famously said, or maybe infamously said, I ship code. I don't read and he's not alone. Right? Most, you know, you you you've seen these stories from, some of the head devs at anthropic. People are just saying, yeah, I don't I don't read the code. Right. Maybe it's they have a lot of confidence in whatever product that they're using because they're also building it. But it's kinda like if a conveyor belt was crate to 10 x speed, and you still just had one quality expect inspector that's just holding a clipboard. Right.
Jordan Wilson [00:28:01]:
There's it's seemingly as the AI tools get better and smarter, more efficient. Well, it seems like, well, we just trust them more. And, when engineering teams can't keep up with the amount of code that spit out, you know, frustrated employees just start building software entirely on their own. Right. And I'm I'm talking about that manually, right. More and more people as we start wanting AI and we know what's easy, some engineering teams that are still, you know, doing AI kind of maybe just code, AI assisted and not totally vibe coding. Well, they're backed up. Right? Because now all of a sudden, company software stacks have doubled or tripled.
Jordan Wilson [00:28:42]:
Right? And what's this led to? Well, it's also IT sprawl. Right? It's it's AI sprawl. So now you have, code going into production made by people who maybe shouldn't be, deploying code live, but this is also an issue. Right? Because engineering teams that are still maybe doing things the quote, unquote older school way or just using AI code assist or AI code completion and AstraZen vibe coding. Right? The demands on them are way more, and, you know, non technical people are just taking this into their own hands. So I also want to talk about this because I think it's important. I do think that vibe coding will, quote unquote destroy traditional software engineering jobs. But I think overall, this is going to be a eventually a growing field because there's gonna be a whole new category of work that did not exist before.
Jordan Wilson [00:29:36]:
And I'm talking about rescue engineering. So here's this kind of, this ninety day, this three month, kind of realization. So engineers are calling this the three month black box effect. This is essentially when an AI built architecture becomes untouchable. Right? This is when the house, right, that you went in and built after ninety days, you're like, wait. I forgot everything about this. No one knows which walls are low buried. The AI forgets how it built the foundation.
Jordan Wilson [00:30:05]:
Everything stalls. Right? So this is what we're seeing from developers. They number one, everyone's creating exponentially more code. And then we have things like humans are forgetting things. I don't know if you've realized this, pre AI. I think I was smart. Right. I'm not saying that I'm dumber now.
Jordan Wilson [00:30:30]:
I just feel I'm less able to retain information because I am now with AI, right? With agentic AI, I'm consuming way more information than I was previously, which means I'm forgetting things faster. And it's seemingly, this is a, kind of turning into a global problem where after ninety or so days, right, software teams are just kind of forgetting, not just that, but then also, yes, over the last couple of weeks, month or so, context windows in these programs have gotten better. Yes. But still, some of the the, you know, apps that were built, you know, six months ago, and it took a while to get it approved, but it's been live. And now people want updates. Okay. Well, we're way past the context window. The human forgot.
Jordan Wilson [00:31:17]:
The AI forgot. Infrastructures changed. Right? And maybe the code was shoddy to begin with. So this has led to a completely new category of work, which is rescue engineering. And I think this is one of the hottest new disciplines in tech. So, studies have showed an estimated 8,000 startups now need full or partial rebuilds or refactors of their entire code base or apps. And those can go anywhere from, you know, 50,000 to 500,000 to a couple million dollars. Right? You have a lot of AI startups that essentially buy coded a very popular platform.
Jordan Wilson [00:31:53]:
They just started stacking duct tape on top of paper clips, on top of glue, and then they just realized, oh my gosh. We just realized that our app, it's wildly popular. It's exploding, and it's built on, Popsicle sticks. Right? And they're having to go through and completely, redo it. So how do you avoid kind of the fake dream house and build a solid foundation with vibe coating? Here is how you can do it correctly. I think first you have to come to the realization that this is permanent. Right. And even if your team may not be vibe coding right now, you will be by working right.
Jordan Wilson [00:32:33]:
Using something like Claude Code, or sorry, Claude Cowork, using something like Microsoft Copilot Cowork, something like the, you know, whatever the codecs chat g p t super app becomes. This is going to become very common where we are going to be using natural language to build little apps, right? Maybe it's to build apps, to finish our work faster. Maybe it's to build apps for consumers. Maybe it's replacing little pieces of software, updating pieces of software, Non technical people are gonna be doing it. So you need to follow these rules. But also you need to understand the company's riding the waves without governance are gonna pay for their rescue engineering in the brief in the breach costs. But the company's treating AI coding as a rigorous engineering, kind of, problem, with human oversight are gonna capture the real value. Because like I said, this technology is not optional, but how your organization governs it is entirely your choice.
Jordan Wilson [00:33:32]:
Alright. I hope this one was helpful. Taking a quick ish look at the vibe coding boom and why vibe coding isn't going away and how it's both good and bad. Alright. Should we do more vibe coding? Alright. Let me know. I'm always on the on the fence about this, right, especially from a podcast perspective. I'd like to have something, for you all that you can just kinda listen to in your car, you know, listen to when you're on the treadmill walking your dog, whatever it is, and not necessarily have to be watching something on the screen at all times.
Jordan Wilson [00:34:03]:
So if we should be covering more on vibe coding, let me know. But if this is helpful, you're definitely gonna love the rest of our start here series. Remember, this is kinda made to listen to in order. They're shorter, episodes. Alright? So make sure you go to starthereseries.com. That's gonna get you free access to our inner circle community, and you can go listen to all of the episodes in order. So thank you for tuning in. Hope to see you back tomorrow and every day for more everyday AI.
Jordan Wilson [00:34:29]:
Thanks, y'all.