Let's start the AI revolution with a joke! "Why was your AI cold?" "Because you left the windows open!". Funny! And now is the time to explore how AI will shape the future of Full-Stack development.

There is a revolution going on! The AI revolution! Let's talk about the opportunities and challenges that the use of AI and particularly Generative AI brings to full-stack development.

Generative AI automates many development tasks, from writing frontend components to configuring CI/CD pipelines.

While AI can generate code, it lacks true understanding of business logic, scalability, and security concerns.

AI-generated code is only as good as the input prompt and requires human oversight for optimisation and security.

The first AI-generated output is often incomplete or incorrect, making human expertise crucial.

So, how can you use AI to build full-stack systems?

Tools like ChatGPT and GitHub Copilot can create React, Vue, Express.js, and Django components.

AI can assist in writing SQL and NoSQL queries, though human validation is required.

AI can automate deployments using GitHub Actions or Jenkins, improving workflow efficiency. So, considering what we learned about full-stack development, AI can cover pretty much everything. Does that mean that hoomanz are no longer needed?

We think otherwise! Consider this AI-generated React component. While it provides an essential starting point, it needs styling, state management, and business logic. Without engineering or domain-specific knowledge, we would not know how to prompt AI systems to add such functionality.

Let's dive into problems you may encounter when using generative systems to create your code.

One of the main problems is security. AI may generate insecure code, such as missing input validation or hardcoded credentials.

Please take a look at this simplistic example. Did you spot the error? Yes, this login API is insecure because it hardcodes credentials and lacks password hashing.

AI also often produces inefficient database queries or architectures that don't scale well. Moreover, AI struggles with large code bases, having a sluggish performance or providing unrelated code changes or feedback.

Last, AI doesn't understand specific company needs, goals or strategies; it generates "generic" solutions. While it is possible to encode such requirements into a prompt, maintaining it introduces unreasonable effort.

Let's take a look at some feedback from developers who tried to use AI to develop production-ready systems.

Mayo is a senior developer who has spent over 1000 hours working with Cursor AI. That is about 125 days of work. He identified three major issues when using AI.

First, the generated output often contains bugs that could cost you a ton of time and money As a result, a lot of time can be wasted reviewing and refactoring Al generated code.

Second, the output generated by AI is largely inconsistent. If you provide the same prompt several times to the chatbot, you may end up with drastically different solutions to the problem. In addition, if you ask the Al any questions that contain suggestions or alternative solutions, it will apologize and refactor the entire code again.

Last, AI can significantly increase technical debt. Even when the Al generates a "good" code block solution, it doesn't consider the entire software design and architecture of the application. Due to the lack of a holistic perspective, problems emerge, including inconsistent error handling, modularity, and data modelling across the app. As a result, the short term quick fix often leads to scalability, performance, and maintenance issues over the long run. Once your codebase has grown to a large, complex web of messy components, refactoring will be a long, painful and costly process.

The Cursor AI (or AI coding in general) is a useful autocompletion tool that can boost your development productivity in the short run, but in the long run, it can waste significant time and energy IF you don't thoroughly review generated outputs. Cursor AI is best used as a "junior developer" who often makes mistakes, and you must review their work carefully to guide them correctly. If you simply "trust" the AI outputs due to lack of knowledge, skill, or willingness to review results, the long-term damage will outweigh the initial productivity gains you got so hyped about.

Where are humans needed when using AI coding systems?

Developers must manually check AI-generated code for vulnerabilities.

AI-generated code is generic; human developers tailor it to real-world needs.

AI makes mistakes—developers ensure efficiency, correctness, and maintainability.

Check out this list of amazing tools that enable AI-powered coding. Please explore them in your own time and reflect on how much effort it takes to build genuine, fully featured, secure systems and not just toy examples these systems come with.

In your own time, please pause this presentation and watch the video, where two experienced engineers enthusiastic about AI technologies use cutting edge AI technology to build a service for upscaling images. When watching, ask yourself the following questions: First, could AI alone build the app, or did developers need to intervene? Second, is the process straightforward and intuitive, or is engineering knowledge still required? Last, Could you solve their challenges?

Let's summarise what we've learnt

Generative AI is a powerful tool, but human developers ensure correctness, security, and scalability.

Never trust AI-generated code blindly—manual checks are essential.

However, developers who master AI-assisted coding while understanding full-stack principles will be in high demand as they will be efficient in their code production while maintaining the high quality of their work.