The rapid spread of artificial intelligence (AI) has fundamentally changed the landscape of software engineering, simultaneously accelerating productivity and exposing serious gaps in developer education. New and novice developers can now use AI tools to instantly create functional user interfaces, database models, and backend routes with minimal effort.
However, this automation has shifted the main technical responsibility from writing simple syntax to applying complex engineering judgment. Developers now face the urgent challenge of evaluating whether the AI-generated logic is correct, whether data is secure, and whether the system is strictly enforcing the underlying business rules.
The developer workforce in Bangladesh is made up of a dual education system. While formal computer science programs continue to provide the necessary academic discipline and theoretical framework, a large informal education system has emerged as a practical bridge to employable skills in industry.
This system relies largely on public tutorials, community events, and local content. In this informal setting, technology educators typically focus on three distinct learning levels: Teaching tool proficiency (operating specific AI tools), project momentum (building and deploying applications quickly), and engineering judgment.
While analyzing the developer education system in Bangladesh, I came across the work of a software educator who specifically focuses on this third and most difficult layer -- engineering judgment.
He highlighted the cultural shift towards an informal reliance on AI, often colloquially known as “vibe coding.” According to his approach, while the medium of instruction has shifted from traditional programming languages to natural language AI prompts, the core responsibility of the engineer remains unchanged.
He argues that a developer must clearly define software requirements, assess architectural integrity, test edge cases, and deeply understand system behaviour. When these rigorous engineering standards are applied, the use of AI is no longer informal “vibe coding” but rather a highly disciplined engineering practice.
To systematize this philosophy, there is a procedural framework called the "AI coding loop." This procedural cycle explicitly rejects the idea of viewing AI as an automated shortcut and instead presents it as an iterative drafting assistant.
This loop forces the developer to set specific procedural goals, establish strict system constraints, provide clear examples, request code generation in small and manageable steps, implement automated unit testing, and rigorously review the output before iterating the process.
He uses real-world case studies to highlight the hidden vulnerabilities inherent in AI-generated code, particularly in the context of e-commerce shopping cart infrastructure. He shows that an AI tool can easily create a frontend interface that looks perfectly functional -- where buttons are clicked and orders are successfully saved to the database -- yet it is completely missing critical backend security protocols.
If a system incorrectly trusts pricing data sent directly from the client side without backend validation, malicious users can exploit browser requests to purchase products at the wrong price.
The true accuracy of software depends on securing these invisible trust boundaries, which are often overlooked by novice developers who rely solely on AI output.
Realizing that modern AI tools require a structured approach to securely access external data layers, there is a need to prioritize building basic mental models and teach the functional definitions and interactions of 'models', 'contexts', and 'protocols' before moving on to technical implementations.
This fundamental approach ensures that students are not just memorizing isolated tools, but are actively connecting new technologies to existing backend service concepts.
While this concept-centric approach may seem slow to beginners who want to get up to speed quickly on projects, it fills an important gap in the informal education system.
In addition, presenting this complex content in Bengali offers distinct educational benefits beyond mere translation. It allows teachers to shine a light on specific conceptual areas where native learners get stuck and helps them use culturally relevant examples to clarify complex structures.
Finally, as AI dramatically reduces the cost of writing code while simultaneously increasing the risk of shallow technical knowledge, specific educational frameworks are essential for training a workforce capable of critically analyzing machine-generated output.
Anupom Sarker is a freelance contributor.