OpenAI publishes work on coding in low-resource programming languages, including SQLite and COBOL migration evaluations
OpenAI published a post describing efforts to improve ChatGPT's performance across low-resource programming languages. The post introduces new evaluations covering software engineering, SQLite, and COBOL migration.
Why it matters
Builders working with legacy or niche languages such as COBOL or with embedded databases like SQLite may see changes in how well ChatGPT handles those tasks, and new public evaluations give a way to measure that performance.
Who should care
Software engineers maintaining legacy systems, teams planning COBOL migrations, developers building on SQLite, and companies evaluating AI coding assistants.
What you can do
Read the full post and test ChatGPT on a small, representative COBOL or SQLite task from your own codebase to benchmark its current output quality.