OpenAI Improves ChatGPT’s Coding Skills in COBOL, Fortran and Visual Basic

Artificial intelligence is becoming an important tool for software developers, helping them write code, identify errors, and improve existing applications.

However, AI coding assistants have generally performed better in popular programming languages such as Python and JavaScript.

This creates a challenge for organisations that depend on older or less commonly used programming languages.

On October 19, 2026, OpenAI published research describing improvements in its models’ ability to work with languages such as COBOL, Fortran, and Visual Basic.

These languages may be less popular among new programmers, but they remain important in banking, scientific computing, and business software.

The improvements could help organisations maintain and modernise older applications.

But what are these programming languages, why are they still used, and how can AI help?

What Are Programming Languages?

A programming language is a system of instructions used to tell computers what to do.

For example, developers use programming languages to create websites, mobile applications, financial systems, and computer programs.

Different languages are designed for different purposes.

Python is widely used in artificial intelligence, data analysis, and automation.

JavaScript is commonly used to create interactive websites.

SQL is used to retrieve and manage information stored in relational databases.

However, many important organisations still use older programming languages that were developed decades ago.

What Are Low-Resource Programming Languages?

OpenAI uses the term low-resource programming languages to describe languages for which AI systems have comparatively limited training material.

AI coding models learn patterns from large amounts of code, documentation, and related technical information.

Popular programming languages often have extensive publicly available examples.

For example, developers frequently share Python and JavaScript code through tutorials, open-source projects, and technical forums.

Other languages have less publicly available training material.

This can make it harder for AI systems to understand unusual syntax, older programming practices, and specialised software environments.

Low-resource does not necessarily mean that a language is obsolete or unimportant.

Some of these languages remain essential to major industries.

What Is COBOL?

COBOL stands for Common Business-Oriented Language.

It was developed in the late 1950s and became widely used in business computing.

COBOL is particularly associated with banking, insurance, government systems, and large business applications.

Many organisations still use COBOL because their systems have been developed and maintained over several decades.

For example, a bank might use a legacy application to process financial transactions.

Replacing the entire system could be expensive and risky.

The application may interact with many other systems, and even a small mistake could affect important services.

As a result, organisations often continue maintaining established COBOL applications rather than replacing them immediately.

What Is Fortran?

Fortran is a programming language traditionally associated with scientific and engineering calculations.

Its name comes from Formula Translation.

Fortran has been used in fields such as weather forecasting, physics, numerical modelling, and scientific simulation.

For example, scientists may use computational models to study atmospheric conditions.

These simulations can involve large numbers of mathematical calculations.

Fortran remains important because it performs well in certain numerical computing applications and supports extensive existing scientific software.

However, maintaining older Fortran code can be challenging when documentation is limited or experienced developers are difficult to find.

What Is Visual Basic?

Visual Basic is a programming language developed by Microsoft.

It became popular for building Windows applications and automating business processes.

Visual Basic for Applications, commonly known as VBA, is also widely associated with Microsoft Office automation.

For example, an Excel user can create a VBA macro to organise information, generate reports, or perform repetitive spreadsheet tasks.

A company may have developed hundreds of macros over many years.

Although newer automation technologies exist, replacing those macros can require considerable work.

Improved AI understanding of Visual Basic and VBA-related code could help developers analyse and maintain some of these applications.

However, the exact effectiveness of AI support depends on the language version and the software environment.

Why Has AI Struggled With Older Programming Languages?

AI coding assistants learn from examples and training data.

When a programming language has extensive documentation and many code examples, an AI model may have more opportunities to learn common patterns.

However, some older business systems use specialised programming conventions.

Their source code may be private and unavailable in public training datasets.

They may also depend on software that is no longer commonly taught.

For example, a COBOL application written decades ago might use conventions that are unfamiliar to developers accustomed to Python.

An AI system could misunderstand the logic or suggest changes that are incompatible with the existing system.

This creates risks when developers use AI to modify important software.

What Has OpenAI Improved?

OpenAI says its latest model advances provide stronger capabilities for creating, maintaining, and migrating software written in low-resource languages.

Migration refers to moving an application or system from one technology environment to another.

For example, an organisation might want to migrate an older application to a modern platform.

AI could assist by explaining the existing code, identifying important business rules, generating tests, and suggesting possible replacements.

OpenAI has also developed evaluations covering software engineering, SQLite, and COBOL migration.

These evaluations help researchers measure how effectively AI systems perform different programming tasks.

However, improved benchmark performance does not guarantee that an AI model can safely modernise every legacy application.

How Could AI Help Modernise Legacy Software?

Legacy software refers to older systems that remain in use.

These systems may be difficult to maintain because they rely on specialised technologies or limited documentation.

AI tools could help in several ways.

1. Explaining Old Code

A developer could ask an AI assistant to explain what a section of COBOL or Fortran code does.

For example, the AI might identify a calculation or explain how a program handles different conditions.

The explanation could help a new developer understand unfamiliar code.

However, the explanation should be checked against the actual program behaviour.

2. Finding Programming Errors

AI may help identify potential mistakes or inconsistencies in older code.

For example, it could highlight a variable that appears to be used incorrectly.

The developer would then investigate whether the issue is genuine.

AI-generated warnings should not automatically be treated as confirmed defects.

3. Generating Tests

Before modifying an old application, developers need to understand its existing behaviour.

Automated tests can help confirm whether the software continues producing expected results after changes.

AI coding assistants can help generate test cases.

However, human developers must ensure the tests cover important business rules and unusual situations.

4. Translating Code

AI tools may assist with converting software from one programming language to another.

For example, a developer could explore how a COBOL routine might be rewritten using a modern programming language.

However, translating syntax is only one part of software migration.

The new system must preserve the original application’s business logic, data integrity, security, and reliability.

This requires extensive testing.

Why Is This Important for Banks?

Banks depend on software systems that process sensitive financial information.

Some of these applications have been operating for decades.

Replacing them can be particularly challenging because failures may affect payments, account balances, or other important services.

AI-assisted tools could help developers understand and document the older systems.

This may support gradual modernisation.

However, banks must carefully control the use of AI when working with confidential information and critical infrastructure.

They also need to verify that any generated code meets security and regulatory requirements.

Could AI Make Older Programming Languages Popular Again?

Improved AI assistance might make it easier for developers to work with older languages.

For example, someone familiar with Python could use an AI assistant to understand an unfamiliar COBOL program.

This could reduce some of the barriers involved in learning legacy technologies.

However, it does not necessarily mean that COBOL or Fortran will become more popular than modern programming languages.

Organisations will continue choosing technologies based on performance, reliability, cost, available skills, and business requirements.

Older languages may remain useful for specific applications while newer languages dominate other areas.

What Does This Mean for African Developers?

Many banks, businesses, and government organisations across Africa use a combination of modern and older software systems.

Developers with skills in databases, business applications, and legacy technologies may find opportunities in maintenance and modernisation projects.

AI tools could help these professionals understand unfamiliar systems and complete some tasks more efficiently.

For example, a developer working with an older financial application could use AI to help explain complicated code while independently verifying its behaviour.

However, organisations should avoid uploading confidential source code or sensitive customer information to AI services without proper authorisation.

Developers must follow applicable security policies and data protection requirements.

Should Beginners Learn COBOL or Fortran?

The answer depends on their career interests.

For someone beginning a general programming journey, Python or JavaScript may offer an accessible introduction.

Python is particularly useful for data analysis, automation, and artificial intelligence.

JavaScript is valuable for web development.

SQL is important for working with databases.

However, COBOL may be relevant for careers involving banking and legacy enterprise systems.

Fortran may be useful in scientific computing, engineering, and numerical research.

Visual Basic and VBA can remain valuable in organisations that depend on Microsoft Office automation.

Beginners should choose languages based on the problems they want to solve rather than popularity alone.

QueryFun Take: Old Programming Languages Still Matter in the AI Era

OpenAI’s research highlights an important reality of software development.

New technologies do not always replace older systems immediately.

Many businesses depend on software created years or even decades ago.

As experienced developers retire and documentation becomes harder to find, maintaining these systems can become increasingly difficult.

AI coding assistants could help preserve technical knowledge and support software modernisation.

However, developers must still understand the underlying logic and test changes carefully.

For students, the lesson is that programming fundamentals are more important than memorising the syntax of one language.

Understanding algorithms, databases, debugging, and software testing makes it easier to work across different technologies.

Frequently Asked Questions

What are low-resource programming languages?

They are programming languages with comparatively limited training data or documentation available to AI models.

Is COBOL still used today?

Yes. COBOL remains important in some banking, government, insurance, and enterprise systems.

What is Fortran used for?

Fortran is widely associated with scientific computing, numerical analysis, engineering, and simulation.

Can ChatGPT understand Visual Basic?

AI coding assistants can help explain and generate Visual Basic code, but accuracy depends on the application and code complexity.

Can AI convert COBOL into Python?

AI can assist with translation and migration, but the resulting software requires extensive testing to preserve functionality.

Is it worth learning VBA?

VBA can still be useful for automating Excel and other Microsoft Office applications, particularly in workplaces with existing macro-based workflows.

Which programming language should beginners learn first?

Python is often a practical starting point, especially for people interested in automation, artificial intelligence, and data analysis.

Primary source: OpenAI: Advancing Coding in Low-Resource Programming Languages