Code generators vs AI Assisted Coding

Code generators vs AI Assisted Coding

The Historical Significance of Delta

Learn Java and get the Oracle OCP Certification

Today, AI assisted coding is changing the way software is designed and developed. Tools can generate code from natural-language instructions, automate repetitive programming tasks, and help developers work at a higher level of abstraction.

But the underlying idea is far from new.

Long before generative AI, there were technologies designed to let programmers describe what an application should do while automatically generating much of the how. One of the most significant examples was Delta, an early industrial-scale COBOL and PL/I code-generation technology developed for enterprise and IBM mainframe environments.

Emerging in the 1970s, Delta was created at a time when large business systems were dominated by COBOL, PL/I and mainframes. Enterprise applications could contain enormous amounts of repetitive code, particularly around data processing, files, databases, transaction processing and business rules. Delta addressed this problem by allowing developers to work at a higher level of abstraction and generate conventional COBOL or PL/I source code.

This was much more than simple text templating. Delta evolved into a comprehensive generative development environment, including reusable generators and macros capable of producing application code for different technical environments. It could help separate business logic from platform-specific implementation, supporting technologies such as IBM mainframes, CICS, IMS and databases including DB2.

The significance of this approach becomes clearer when viewed through a modern lens:

Delta was an early, industrial-scale COBOL and PL/I code generator that demonstrated decades before generative AI that software could be created by describing applications at a higher level and automatically generating implementation code.

That philosophy has remarkable parallels with today's AI assisted coding.

The technologies are obviously very different. Delta relied on explicitly defined generators, schemas, macros and deterministic rules. Modern AI coding systems use large language models capable of generating code from natural-language prompts and existing codebases. Delta generally produced predictable output from structured input; AI-generated code can be probabilistic and requires human validation.

Yet the fundamental ambition is similar: reduce the amount of repetitive low-level programming that humans have to write and maintain.

Delta also introduced an important concept that remains relevant today: the generated source code was not necessarily the true development artifact. In a Delta environment, the higher-level Delta definitions, macros and generator configuration could be considered the real source, while the resulting COBOL or PL/I was the generated implementation.

That distinction is increasingly relevant in the age of AI. As developers rely more heavily on generated code, the important question is not simply “Who wrote this code?” but “What higher-level specification, intent or process produced it, and can we reproduce and understand that process?”

Seen from today's perspective, Delta therefore represents an important chapter in the history of software automation.

AI assisted coding did not invent the idea of generating software from higher-level descriptions. Delta and other early code-generation technologies were already exploring that vision decades ago—on an industrial scale, inside some of the world's most critical enterprise systems.

Learn Java and get the Oracle OCP Certification

pcworkshopslondon.co.uk

Comments

Popular posts from this blog

Building Cross-Platform JavaFX Apps with Gluon

Companies and real-world applications using JavaFX

Delete vs Truncate in MySQL and MS SQL Server