Absolutely true.

[bobthomson70@mastodon.social]

@cstross I saw a comment on the Reg story on this that made again the point that often gets missed with the talk of replacing COBOL, that it's not just the COBOL, and in many ways that's the easiest part, it's the surrounding transaction layers and databases and z/OS filesystem components, most of which don't really map onto the usual Linux based platforms very easily at all.

[marfymazie@yiff.life]

@skolima @cstross I second that I don't think finance is conservative for tech. They can be quite bullish on tech, I've worked with a small one who was very into Blockchain and early AI swell (they don't exist anymore for obvious reasons). Just look at the push to use only app, no in person service, they are one of the biggest drivers of it.
Law firms are arguably far more fuddy and slow to move forward (having just worked for one, they had staff who still hand wrote or dictated for transcription by typists). Some have been caught making stuff up for court because an AI made up a case and law etc.
AI is going to be in your bank, because their investment arms are buying into it and the bubble must grow (IMHO)

[jsl@hachyderm.io]

@fn0rd @cstross No, the old rules still apply the moment you start fiddling around with the government's flow of money. Banks pump the money around. You don't change the core infrastructure just because a few hyperscalers offer your IT teams a few million $$$ of free AI tokens.
What you do do is that you may try LLMs in the customer interface, maybe in fraud detection and perhaps in documenting your code.
There's no way that the next flash crash will be blamed on a few interns who vibe coded a major bank's e-trading core or transaction banking engine.
The same may not apply to challenger banks or shadow banks, though.

[fn0rd@mastodon.social]

@jsl @cstross
> the old rules still apply the moment you start fiddling around with the government's flow of money.

It is almost inconceivable to me that this is the *optimistic* way to look at this issue :)
But politicians seem to be pretty cheap and easy to buy these days, so while I hope you are correct, I don't share your optimism.

[mortonrobd@mas.to]

@bobthomson70 @cstross can confirm. Plus the programmers who could explain how the mainframe code works are all busy last I heard addressing genuine fiduciary issues to explain the code. Plus the code interactions are insane - cross cutting concerns all over.

[cstross@wandering.shop]

@MarfyMazie @skolima

The only app/no in-person in banking is a cost-control thing: app-mediated service is *vastly* cheaper than putting trained staff behind a counter in prime high street real estate (with banking security on top).

Law firms deal with processes that have time scales measured in decades to (occasionally) centuries. Banks don't usually have to worry about anything older than the term of a just-now-maturing mortgage. Trusts, however …

[causticmsngo@mastodon.social]

@jawarajabbi Apologies; I was responding to the hope that “people who run banks would never do something risky” like using LLMs in their systems because someone might go to jail if they screw it up. Most of my experience working with banks & investment companies is with large ones. I think they absolutely would & no one would face any real consequences if that lead to a big screw up.

[larsmb@mastodon.online]

@etchedpixels @cstross This fits my previous experiences in such regulated and "conservative" IT projects.

However, I *am* concerned that regulators will look the other way or lower restrictions to sustain this bubble. (The US has already dropped a few safeguards across the entire spectrum, after all.)

And if they do, the finance sector *will* jump on this, because if anyone wants to pinch pennies at scale, it's them.

[thepolishdispatch@mstdn.social]

@cstross Central-European-centric tangent: average USian hasn't got the slightest clue how far behind US banking is relative to what we have in central-european countries, formerly behind the Iron Courtain. We did not have any ancient ballast dragging us down, hence, all computerized banking systems are fairly modern (think technologies available after 1989). Websites, applications, functionalities, speed, low or nonexistent fees, ...

[dukeduke@mastodon.social]

@cstross Just one more good reason to only use banks as minimally as possible, and NEVER keep your savings with them. And that includes safety deposit boxes. If you don't hold it--you don't own it.

[danmcd@hostux.social]

@bobthomson70 @cstross

Don't dismiss i-Series (nee AS/400) either, as well as its near-native RPG!

[medeavanamonde@beige.party]

@skolima @cstross

It’s been a generation at least and no one remembers the S & L Bank crash

[paradegrotesque@mastodon.sdf.org]

@MortonRobD @bobthomson70 @cstross

I... Did work with people who were very busy porting z/os cobol code to Linux + Oracle + NetCobol...

So I know it can be done. Sorta.

It was... painful. Old cobol programmers tend to be crusty fellows, who don't really like these new 'PC' things.

[thepolishdispatch@mstdn.social]

@cstross If it wasn't because of different design and architecture (e.g. IBAN vs. ABA), the simplest, fastest, and most cost-effective path to modernizing US banking would be for them being taken over by European ones. (granted, I have no experience with western-european ones, but I can vouch for those central-european).