Who would've thought a sector with gold flowing through its hands would be so stingy when it comes to updating their backend that they'd end up relying on a dying language, and call upon AI to update it for them rather than just paying a competent team to create and rigorously test a new backend in a modern language

[–] aksdb@feddit.de 2 points 11 months ago

One problem is that they need to put a price tag and therefore a timeline on such a project. Due to the complexity and the many unknown unknowns in theses decades worth of accumulated technical debts, no one can properly estimate that. And so these projects never get off and typically die during planning/evaluation when both numbers (cost and time) climb higher and higher the longer people think about it.

IMO a solution would be to do it iteratively with a small team and just finish whenever. Upside: you have people who know the system inside-out at hand all the time should something come up. Downside of course is that you have effectively no meaningful reporting on when this thing is finished.