COBOL devs are becoming migration auditors.

We’ve heard “this tool kills programmers” since compilers, 4GLs, and offshore outsourcing. What it usually kills is the task, not the people — they move one abstraction layer up. The question worth watching is whether this time the layer they move to is big enough to absorb them.

OpenAI’s new models are getting scarily good at modernizing legacy code — and that’s a big deal for finance.

COBOL, Fortran, and Visual Basic still run enormous swaths of the global financial system, and migrating them has traditionally been slow, expensive, and error-prone. OpenAI’s latest benchmark on COBOL migration shows how quickly the gap is closing: GPT-6.1 Sol hits roughly 97% accuracy at just $0.50 per graded trial, while the previous GPT-6 Sol needed about $2 to reach similar performance — a 4x cost reduction in one generation.

OpenAI’s framing is notable too: rather than treating this as a novelty, they’re positioning it as core enterprise capability — understanding decades of buried business logic while preserving what works. They’re also actively soliciting enterprise input on which languages and workflows to tackle next, which suggests this is a strategic priority, not a side project.

#ArtificialIntelligence #COBOL #LegacySystems #FutureOfWork

Learn how OpenAI is improving ChatGPT across low-resource programming languages, with new evaluations for software engineering, SQLite, and COBOL migration.