August 3, 2026
9 min read
Delivery Coordination Manager, Comsys
There is a moment every technology leader recognizes. A developer — or sometimes a product manager with access to the right AI tool — demos a working application they built over a weekend. It looks impressive. It does most of what the business asked for. And it was built in a fraction of the time a traditional development cycle would have required.
That moment is real, and the tools behind it are genuinely powerful. AI coding assistants like Claude Code, Cursor, GitHub Copilot, and Replit Agent have fundamentally changed what is possible for individual builders. The speed is not hype. The productivity gains are measurable.
But there is a question that rarely gets asked in that demo meeting: who is going to run this in production, maintain it, secure it, and adapt it when the business changes next quarter?
That question is where the conversation about vibe coding ends and the conversation about low-code begins. And for most organizations, that question has a clear answer.
The term was coined by Andrej Karpathy in February 2025 to describe a mode of development where you describe what you want in natural language and let an AI generate the code — without necessarily reviewing or fully understanding what was produced. Collins English Dictionary named it Word of the Year for 2025. The speed is the point. The code is almost secondary. [1]
In practice, vibe coding sits on a spectrum. At one end is pure prompt-and-accept development. At the other is what thoughtful practitioners call augmented coding — using AI to accelerate work while reviewing, testing, and genuinely owning the output. The distinction matters enormously in an enterprise context, and it is one that most vibe coding conversations conveniently skip over.
The tools — Claude Code, Cursor, GitHub Copilot Agents, Replit Agent, v0, Lovable — generate real, portable source code in standard languages. You own what is produced. That is a legitimate advantage. The question is whether ownership alone is sufficient for what your organization actually needs.
Speed of creation is only one dimension of enterprise software. The others — security, governance, compliance, maintainability, integration, lifecycle management — are where vibe coding consistently falls short, and where the gap between a promising prototype and a production-ready system becomes expensive.
The security picture alone should give enterprise leaders pause. Independent studies now consistently find that 68–73% of AI-generated code contains security vulnerabilities. Veracode's 2025 GenAI Code Security Report found AI tools failed to protect against cross-site scripting 86% of the time. Stanford's analysis found vibe-coded commits introduce security findings at ten times the rate of human-written code. [2] These are not edge cases — they are the baseline behavior of current AI code generation when left ungoverned.
Beyond security, there is what researchers call comprehension debt. When a function is generated in seconds, the developer sees the output but does not necessarily internalize the logic, edge cases, or dependencies. That debt compounds quietly — slowing incident response, making maintenance harder, and gradually eroding the productivity gains that motivated AI adoption in the first place.
| Vibe coding optimizes for the moment of creation. Enterprise software needs to survive far beyond that moment. |
None of this is an argument against AI-assisted development. It is an argument for being clear-eyed about what problem you are actually trying to solve. Vibe coding is a powerful tool for individual developers, technical prototyping, and exploring ideas. It is not an organizational development strategy.
Low-code platforms — OutSystems, Mendix, ServiceNow App Engine, Salesforce Platform, and enterprise-grade platforms like Jaggle — are development environments built around a fundamentally different set of priorities. Not just speed of creation, but reliability of operation.
The operational advantages are structural, not incidental:
The scale of adoption reflects this. Gartner estimates the low-code market will reach $44.5 billion in 2026, growing at 19% annually, with 75% of large enterprises already using at least four different low-code tools. [3] These organizations are not experimenting. They are standardizing.
Low-code does have real constraints — platform ceilings for highly specialized logic, vendor lock-in considerations that warrant careful TCO modelling, and licensing costs that scale with usage. These deserve scrutiny before any platform commitment. But they are manageable challenges, not fundamental limitations — and as the next section makes clear, the most significant constraint that critics have historically raised against low-code has already been addressed.
The most important development in this space is not vibe coding. It is what happens when AI is embedded directly inside a low-code platform.
Leading platforms are no longer purely visual drag-and-drop environments. They have integrated AI assistants, generative features, and natural language interfaces into the core development experience — enabling builders to describe what they want and have the platform generate flows, interfaces, business logic, and components automatically. The output lands inside a governed, structured, enterprise-ready environment from the moment it is created.
| This is the answer to the false choice between speed and governance. AI-powered low-code delivers both — and it does so within an operational framework that enterprises can actually trust. |
Jaggle, Comsys's enterprise low-code application generation platform, is built on exactly this model. A user describes what they need — a business application, a customer-facing portal, an internal workflow tool — and Jaggle generates it, including React-based interfaces and web pages, directly from natural language input. The platform handles the structure, governance, and lifecycle management automatically.
What makes this genuinely different from vibe coding is the layer underneath. The generated application does not exist in isolation — it inherits enterprise integration, access controls, and deployment infrastructure from the platform. And for teams with the technical capability to go deeper, the generated code is accessible and modifiable. You get the speed of AI generation and the control of traditional development, without having to choose between them.
This directly answers the oldest criticism of low-code: that it is limiting for complex or custom requirements. When AI can generate custom logic and interfaces from a prompt — within a governed platform — the ceiling that once defined low-code's boundaries moves significantly higher. The argument that vibe coding wins on flexibility becomes much harder to sustain.
The question is not which approach is better in the abstract. It is which one is appropriate for what you are actually trying to build.
Vibe coding makes sense when:
Low-code — and specifically AI-powered low-code — is the right choice when:
For most enterprise software — internal tools, business process applications, customer-facing portals, workflow automation, compliance-critical systems — the second list describes the reality. Vibe coding is a developer's instrument. Low-code is an organizational capability.
| Vibe coding helps developers build software faster. Low-code helps organizations build software they can run, govern, maintain, and continuously improve — at scale, with confidence. |
These are not the same problem. The enterprise software graveyard is full of things that were fast to build and expensive to operate. The productivity gains of AI coding tools are real — but productivity in creation is only one part of a software lifecycle that includes deployment, security, maintenance, compliance, and organizational ownership.
The organizations getting the most durable value from technology investment right now are not the ones choosing between vibe coding and low-code. They are the ones who understand which layer each one belongs to — and who are building their development strategy accordingly.
Vibe coding will continue to evolve, and it will remain a genuinely valuable capability for developers and technical builders. No serious technology leader should dismiss it.
But for organizations that need to deliver applications their teams can depend on — that are secure by design, compliant by default, maintainable without specialist intervention, and integrated with the systems the business already runs on — low-code is not a compromise. It is the more mature, more operationally intelligent choice.
And with AI now embedded directly into platforms like Jaggle, the historical trade-off between low-code's governance and vibe coding's speed no longer holds. The question for most organizations is not whether to choose one over the other. It is whether they are ready to build on a platform that gives them both.
References
[1] Karpathy, A. (February 2025). Original vibe coding post. Collins English Dictionary Word of the Year, 2025.
[2] Georgetown CSET study (2025); Veracode GenAI Code Security Report (2025); Stanford analysis of vibe-coded commits. Cited in source document.
[3] Gartner Low-Code Market Forecast, 2026. Cited in source document.
About the Author
Christos Karakasidis is an Information Technology and business professional with more than 18 years of experience in aligning business requirements with technology solutions across the private and public sectors.
Throughout his career, he has focused on coordinating the effective development of technology projects, fostering collaboration between business and technical teams, and supporting operational excellence. He is particularly passionate about mentoring talent and believes that meaningful progress is achieved through empathy, innovation and action.
He is the Delivery Coordination Manager of Comsys, responsible for coordinating project development activities and supporting the successful implementation of technology solutions..
Delivery Coordination Manager, Comsys
There is a moment every technology leader recognizes. A developer — or sometimes a product manager with access to the right AI tool — demos a working application they built over a weekend. It looks impressive. It does most of what the business asked for. And it was built in a fraction of the time a traditional development cycle would have required.
That moment is real, and the tools behind it are genuinely powerful. AI coding assistants like Claude Code, Cursor, GitHub Copilot, and Replit Agent have fundamentally changed what is possible for individual builders. The speed is not hype. The productivity gains are measurable.
But there is a question that rarely gets asked in that demo meeting: who is going to run this in production, maintain it, secure it, and adapt it when the business changes next quarter?
That question is where the conversation about vibe coding ends and the conversation about low-code begins. And for most organizations, that question has a clear answer.
The term was coined by Andrej Karpathy in February 2025 to describe a mode of development where you describe what you want in natural language and let an AI generate the code — without necessarily reviewing or fully understanding what was produced. Collins English Dictionary named it Word of the Year for 2025. The speed is the point. The code is almost secondary. [1]
In practice, vibe coding sits on a spectrum. At one end is pure prompt-and-accept development. At the other is what thoughtful practitioners call augmented coding — using AI to accelerate work while reviewing, testing, and genuinely owning the output. The distinction matters enormously in an enterprise context, and it is one that most vibe coding conversations conveniently skip over.
The tools — Claude Code, Cursor, GitHub Copilot Agents, Replit Agent, v0, Lovable — generate real, portable source code in standard languages. You own what is produced. That is a legitimate advantage. The question is whether ownership alone is sufficient for what your organization actually needs.
Speed of creation is only one dimension of enterprise software. The others — security, governance, compliance, maintainability, integration, lifecycle management — are where vibe coding consistently falls short, and where the gap between a promising prototype and a production-ready system becomes expensive.
The security picture alone should give enterprise leaders pause. Independent studies now consistently find that 68–73% of AI-generated code contains security vulnerabilities. Veracode’s 2025 GenAI Code Security Report found AI tools failed to protect against cross-site scripting 86% of the time. Stanford’s analysis found vibe-coded commits introduce security findings at ten times the rate of human-written code. [2] These are not edge cases — they are the baseline behavior of current AI code generation when left ungoverned.
Beyond security, there is what researchers call comprehension debt. When a function is generated in seconds, the developer sees the output but does not necessarily internalize the logic, edge cases, or dependencies. That debt compounds quietly — slowing incident response, making maintenance harder, and gradually eroding the productivity gains that motivated AI adoption in the first place.
| Vibe coding optimizes for the moment of creation. Enterprise software needs to survive far beyond that moment. |
None of this is an argument against AI-assisted development. It is an argument for being clear-eyed about what problem you are actually trying to solve. Vibe coding is a powerful tool for individual developers, technical prototyping, and exploring ideas. It is not an organizational development strategy.
Low-code platforms — OutSystems, Mendix, ServiceNow App Engine, Salesforce Platform, and enterprise-grade platforms like Jaggle — are development environments built around a fundamentally different set of priorities. Not just speed of creation, but reliability of operation.
The operational advantages are structural, not incidental:
The scale of adoption reflects this. Gartner estimates the low-code market will reach $44.5 billion in 2026, growing at 19% annually, with 75% of large enterprises already using at least four different low-code tools. [3] These organizations are not experimenting. They are standardizing.
Low-code does have real constraints — platform ceilings for highly specialized logic, vendor lock-in considerations that warrant careful TCO modelling, and licensing costs that scale with usage. These deserve scrutiny before any platform commitment. But they are manageable challenges, not fundamental limitations — and as the next section makes clear, the most significant constraint that critics have historically raised against low-code has already been addressed.
The most important development in this space is not vibe coding. It is what happens when AI is embedded directly inside a low-code platform.
Leading platforms are no longer purely visual drag-and-drop environments. They have integrated AI assistants, generative features, and natural language interfaces into the core development experience — enabling builders to describe what they want and have the platform generate flows, interfaces, business logic, and components automatically. The output lands inside a governed, structured, enterprise-ready environment from the moment it is created.
| This is the answer to the false choice between speed and governance. AI-powered low-code delivers both — and it does so within an operational framework that enterprises can actually trust. |
Jaggle, Comsys’s enterprise low-code application generation platform, is built on exactly this model. A user describes what they need — a business application, a customer-facing portal, an internal workflow tool — and Jaggle generates it, including React-based interfaces and web pages, directly from natural language input. The platform handles the structure, governance, and lifecycle management automatically.
What makes this genuinely different from vibe coding is the layer underneath. The generated application does not exist in isolation — it inherits enterprise integration, access controls, and deployment infrastructure from the platform. And for teams with the technical capability to go deeper, the generated code is accessible and modifiable. You get the speed of AI generation and the control of traditional development, without having to choose between them.
This directly answers the oldest criticism of low-code: that it is limiting for complex or custom requirements. When AI can generate custom logic and interfaces from a prompt — within a governed platform — the ceiling that once defined low-code’s boundaries moves significantly higher. The argument that vibe coding wins on flexibility becomes much harder to sustain.
The question is not which approach is better in the abstract. It is which one is appropriate for what you are actually trying to build.
Vibe coding makes sense when:
Low-code — and specifically AI-powered low-code — is the right choice when:
For most enterprise software — internal tools, business process applications, customer-facing portals, workflow automation, compliance-critical systems — the second list describes the reality. Vibe coding is a developer’s instrument. Low-code is an organizational capability.
| Vibe coding helps developers build software faster. Low-code helps organizations build software they can run, govern, maintain, and continuously improve — at scale, with confidence. |
These are not the same problem. The enterprise software graveyard is full of things that were fast to build and expensive to operate. The productivity gains of AI coding tools are real — but productivity in creation is only one part of a software lifecycle that includes deployment, security, maintenance, compliance, and organizational ownership.
The organizations getting the most durable value from technology investment right now are not the ones choosing between vibe coding and low-code. They are the ones who understand which layer each one belongs to — and who are building their development strategy accordingly.
Vibe coding will continue to evolve, and it will remain a genuinely valuable capability for developers and technical builders. No serious technology leader should dismiss it.
But for organizations that need to deliver applications their teams can depend on — that are secure by design, compliant by default, maintainable without specialist intervention, and integrated with the systems the business already runs on — low-code is not a compromise. It is the more mature, more operationally intelligent choice.
And with AI now embedded directly into platforms like Jaggle, the historical trade-off between low-code’s governance and vibe coding’s speed no longer holds. The question for most organizations is not whether to choose one over the other. It is whether they are ready to build on a platform that gives them both.
References
[1] Karpathy, A. (February 2025). Original vibe coding post. Collins English Dictionary Word of the Year, 2025.
[2] Georgetown CSET study (2025); Veracode GenAI Code Security Report (2025); Stanford analysis of vibe-coded commits. Cited in source document.
[3] Gartner Low-Code Market Forecast, 2026. Cited in source document.
About the Author
Christos Karakasidis is an Information Technology and business professional with more than 18 years of experience in aligning business requirements with technology solutions across the private and public sectors.
Throughout his career, he has focused on coordinating the effective development of technology projects, fostering collaboration between business and technical teams, and supporting operational excellence. He is particularly passionate about mentoring talent and believes that meaningful progress is achieved through empathy, innovation and action.
He is the Delivery Coordination Manager of Comsys, responsible for coordinating project development activities and supporting the successful implementation of technology solutions..
