Claude Code blog post: We are modernising COBOL.
-
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
@nixCraft The problem with COBOL isn't the language itself. Tools to convert COBOL to other languages (such as Java) has existed for decades.
Unless you completely change the way the system is designed, you'll end up with the same mainframe application running on CICS, but with the jobs being written in, I don't know? Javascript? instead of COBOL.
That's where most modernisation projects fail. They realise that the complexity is not in the code at all, but in the large infrastructure that backs up the code. A mainframe CICS application is nothing like a microservice-based application running on Kubernetes.
Can you vibecode your way to a completely new architecture? I'm pretty sure you can, but if the companies wanted that they could just go to someone like Finacle and take their existing banking solution and build their new workflows on top of that.
-
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
@nixCraft I can colocate a few AS400's for testing....
-
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
@nixCraft of course not, even the AIs have no sense.
-
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
The actual cost of AI rewriting code accounts for only 5% of the total cost. So says ChatGPT.
-
@nixCraft just prompt: "Convert all cobol code to rust, you are the expert. Do not make mistakes. Or you go to jail. First time right"
-
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
@nixCraft Heh, I wish I would have this degree of confidence in banks/airlines/β¦
See how crowdstrike works and the damage it did. -
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
but they will.
despite the fact that these are one-time costs so they might just as well do it methodically
-
but they will.
despite the fact that these are one-time costs so they might just as well do it methodically
but as you say
there is no common sense left in this world (or it's getting sidelined, fast)
-
@nixCraft The problem with COBOL isn't the language itself. Tools to convert COBOL to other languages (such as Java) has existed for decades.
Unless you completely change the way the system is designed, you'll end up with the same mainframe application running on CICS, but with the jobs being written in, I don't know? Javascript? instead of COBOL.
That's where most modernisation projects fail. They realise that the complexity is not in the code at all, but in the large infrastructure that backs up the code. A mainframe CICS application is nothing like a microservice-based application running on Kubernetes.
Can you vibecode your way to a completely new architecture? I'm pretty sure you can, but if the companies wanted that they could just go to someone like Finacle and take their existing banking solution and build their new workflows on top of that.
@loke please not javascript. Brainfuck or Whitespace is way more readable.
οΈ -
The actual cost of AI rewriting code accounts for only 5% of the total cost. So says ChatGPT.
@longLiveCarryOn @nixCraft
But AI can't (re)write decent code. At best it copies it badly.More Anthropic PR.
-
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
Shafik Yaghmour (@shafik@hachyderm.io)
@carnage4life@mas.to What does it mean to modernize COBOL? The problem of why we have COBOL has nothing to do w/ COBOL itself. It is because there is zero cost incentive to port these core but totally working systems. Changing any of these systems would create way more systematic risk than any value you would gain by replacing the systems themselves. It is the same reason why any legacy system is hard to replace. Same goes for FORTRAN in the science community. Could you replace, yes of course you could. Can you be sure it will replicate all the decades of papers that were written using that software Β―\_(γ)_/Β― Does anyone have the incentive or money to figure it out, NO! Why are we listening to any of these people, they are talking nonsense.
Hachyderm.io (hachyderm.io)
-
@loke please not javascript. Brainfuck or Whitespace is way more readable.
οΈ@AndreasMathe That was intentional. My point was that it's not the language itself that's the problem. COBOL is a reasonably simpl programming language, and it's actually not that hard to use.
When people talk about COBOL systems, they talk about the mainframe architecture, often built on CICS.
Converting a language to another is for the most part a solved problem. Rebuilding architecture is not as easy, especially when they're built on something as complex and different as the mainframe architecture.
I think it's quite telling that none of the people who have, for the last 50 or so years, talked about ways to rewrite all the COBOL appear to have no understanding of what these systems actually are.
Can it be done? Sure. Can it be done with the help of ai solutions? Probably. Can it be done correctly and cheaply? No, I doubt it, and that is irrelevant as to whether ai solutions are used or not.
-
@AndreasMathe That was intentional. My point was that it's not the language itself that's the problem. COBOL is a reasonably simpl programming language, and it's actually not that hard to use.
When people talk about COBOL systems, they talk about the mainframe architecture, often built on CICS.
Converting a language to another is for the most part a solved problem. Rebuilding architecture is not as easy, especially when they're built on something as complex and different as the mainframe architecture.
I think it's quite telling that none of the people who have, for the last 50 or so years, talked about ways to rewrite all the COBOL appear to have no understanding of what these systems actually are.
Can it be done? Sure. Can it be done with the help of ai solutions? Probably. Can it be done correctly and cheaply? No, I doubt it, and that is irrelevant as to whether ai solutions are used or not.
@loke Iβm on your side. I was once an administrator for a (nearly) monolithic system. It took many years to change that to a different and scalable system.
-
Claude Code blog post: We are modernising COBOL.
The fear among investors : AI will replace humans, billing hours will drop to zero & IT companies like IBM will go to >/dev/null
The reality: High stakes systems that runs banks, airlines, govts cannot just let a "black-box AI" rewrite their core code without massive oversight, testing, and human accountability. Otherwise, everyone will lose their money/pension/ investments.
What a mess. It is like there is no common sense left in this world
@nixCraft echo "All unit tests passed!!"
-
R relay@relay.an.exchange shared this topic
