With the proliferation of AI development tools like Claude Code and GitHub Copilot, software development productivity has significantly improved. AI-specific techniques and knowledge, such as context preparation methods and prompt engineering, are being actively shared.
However, now that AI development tools have become widespread, what truly matters in system development is selecting and enforcing effective principles for projects, and utilizing fundamental technologies such as static analysis to detect defects.
Changes Brought by AI Development Tools
AI development tools are undeniably powerful. Claude Code can implement complex logic in a short time, and GitHub Copilot quickly generates repetitive code. However, these tools assist with "how to build" rather than "what to build" .
The Importance of Fundamental Technologies
The success of a project requires selecting appropriate principles. For example:
- Design Principles: SOLID principles, DRY principle, YAGNI principle
- Architecture Patterns: Layered architecture, microservices, domain-driven design
- Security Principles: Principle of least privilege, defense in depth, secure by design
AI tools do not automatically apply these principles. Developers have the responsibility to select principles suitable for the project and enforce them across the team.
In the financial trading system I worked on, we adopted a design for multi-tenant architecture that controls customer boundaries at the infrastructure level, independent of application-specific implementations:
- Strict data isolation using PostgreSQL Row Level Security (RLS)
- Dynamic access control using AWS Dynamic Policy
- IAM role design based on the principle of least privilege
With this architectural design, we built a system with mechanical guardrails that are thoroughly enforced. This mechanism was designed to prevent implementation errors in code written by humans in team development, but it remains equally effective in the AI era. For example, even if a developer mistakenly implements code that attempts to access another customer's data:
-- Example of erroneous implementation
SELECT * FROM customer_transactions WHERE customer_id = 'customer_123';Row Level Security (RLS) ensures that only records associated with the current session's customer are retrieved, preventing access to other customers' data. The system operates safely at the infrastructure level, independent of application code implementation errors.
By establishing mechanical guardrails at the infrastructure level, we can build robust systems that do not depend on the quality of code generated by humans or AI.
The Role of Developers in the AI Era
The role of developers is changing with the advent of AI development tools. However, the essence remains unchanged:
Responsibilities Developers Should Bear
- Architecture Design: Selecting design principles and patterns appropriate for the project
- Enforcing Principles: Formulating and enforcing rules that the entire team should follow
- Systematizing Quality Assurance: Setting up static analysis, test automation, and CI/CD
- Ensuring Security: Conducting vulnerability assessments and security reviews
- Managing Technical Debt: Refactoring and maintaining documentation
AI tools assist with implementation, but they do not replace judgment.
Conclusion
AI development tools are powerful, but they alone do not guarantee project success. Fundamental technologies such as selecting and enforcing effective principles for projects, and quality assurance through static analysis remain critically important in the AI era.
In fact, as the volume of AI-generated code increases, the importance of these fundamental technologies becomes even greater. Effectively utilizing AI tools while ensuring quality through fundamental technologiesâonly with both wheels can sustainable software development be achieved.
References & Related Links
- PostgreSQL Row Security Policies: https://www.postgresql.org/docs/13/ddl-rowsecurity.html