What Happened
On October 19, 2026, OpenAI described work to improve ChatGPT’s capabilities in low-resource programming languages. It identified new evaluations for software engineering tasks and SQLite, with COBOL migration as another evaluation focus. The announcement did not specify a model version, API change, release date, or measured improvement. [1]
Why It Matters to Businesses
Teams maintaining legacy systems need evidence that AI-generated changes preserve business behavior, not just that code compiles. Evaluations focused on COBOL migration and software engineering tasks are relevant, but the announcement alone does not establish that ChatGPT can safely migrate a particular application. [1]
Kimbodo Engineering Perspective
We would treat improved language coverage as a reason to run a controlled pilot, not to relax review. Migration quality depends on application-specific rules, data layouts, database behavior, and operational constraints that a general evaluation may not capture.
How We Would Implement It
- Choose a bounded COBOL component and record its inputs, outputs, dependencies, and current behavior.
- Build regression tests from production-like, de-identified cases, including edge cases and SQLite interactions where applicable.
- Use ChatGPT to propose changes in isolated branches; require compilation, automated tests, semantic comparisons, and human approval before deployment.
- Track defects and review effort against a non-AI migration baseline to determine whether the pilot saves time.
Risks, Costs and Security
The main risk is a change that passes superficial tests but alters financial or operational logic. Legacy code and data may also contain sensitive information. Restrict what is sent to the model, use least-privilege access, retain audit trails, and budget for test construction and expert review. Because no version or performance figures were supplied, buyers should validate the capability in their own environment before planning around it. [1]
Where Kimbodo Comes In
Kimbodo builds and operates this in production for businesses — see our AI Application Development practice, or Estimate My AI Application.
