Holy shit.

Just wow, wow, holy shit:

Completely rewriting a multi-million line COBOL codebase that has life-or-death consequences for real people in the space of a few months, using gen AI?

I’ve been writing software for 40-some years, and I have to say: this may be, without exaggeration, the stupidest software-related idea I’ve ever heard from leadership.

https://www.wired.com/story/doge-rebuild-social-security-administration-cobol-benefits/

@inthehands @alice

I think that characterizing COBOL as 'archaic' is not fair.

Yes, it's not the kind of modern language that most people program in - but it remains fit for the very specific purpose for which it was invented: extremely reliable, consistent, code that remains functional with little to no maintenance for years - decades! - at a time.

Reliability is a desirable quality in infrastructural codebases.

Changing it to something else merely because it's old has no merit - it's creating unnecessary work, and the end result will absolutely not be as reliable as the system it is replacing.

I urge you to listen to the riveting and wonderful radio feature about what happens when you try to build a make-or-break project on a ridiculously tight timeline:

https://99percentinvisible.org/episode/the-worst-video-game-ever/

The page has summary text, but listen if you can. I’ve heard very few instances of a developer speaking so candidly and with such self-awareness about what it feels like to be part of a failing software project.

Listen. Then imagine that instead of E.T. the Extra-Terrestrial, it’s social security for the entire USA.

I always give that story as an assignment to my Software Design and Development students. One of the things we talk about is that if the developer •hadn’t• managed to build the game, if the project had collapsed at any point before release, it probably would have saved the company.

The worst possible outcome here is that these DOGEbags manage to build •something• and actually think it works.

Two things that are not mutually exclusive:

- These people don’t care who they hurt, and would view creating chaos and destroying Social Security as a success.

- These people are so reckless, arrogant, and stupid that they actually think they can do this.

@inthehands

I would caution against calling it the stupidest thing ever. They might take that as a challenge 🙄

It took fifteen years to destroy the Sui Dynasty by this method. We may be witnessing a new world record! Neo-Assyrians did it in 8; they built Babylon on the ruins.

@munin @inthehands @alice @nullpo heck you can do async in modern COBOL, it’s not that outdated of a language really

@q @inthehands @alice @nullpo

.......ok is there documentation on this readily available because "asynch cobol" sounds just delightful tbh

@inthehands if it was as easy as those #CryptoBros / #TechBros think, do they think people like #GraceHopper would've been pulled from retirement trice?

@teledyn @inthehands It'll be fine if they just put a blockchain on it! /s

@munin @inthehands @alice @nullpo it’s in the ISO spec, lemme find a copy

@munin @inthehands @alice @nullpo page 726 https://paperless.magicalcodewit.ch/share/WKJ56gojR7h9scEkWjLU0DfDOrIE4JOTZu1ETI0MpwtN8Gh1zK

@munin @inthehands @alice it's not just fair, it's absolutely FALSE and misleading.

I am very much an expert in the topic, yes. Very, VERY much so.

COBOL is a dead reliable language, and anything 20 years old, means EVERYTHING about it is fully understood. There are no surprises, no unknowns, no failures.

Which is why companies still make COBOL compilers for things like POWER9, z16+, and Red Hat. Even though z16 is fully backwards compatible.

@munin @inthehands @alice it's also a fundamental misunderstanding that COBOL was intended to require little to no maintenance.
COBOL's currently at IEC 1989:2023. Yes, 2023.
The real intent was a portable, data processing language, that was *comprehensible* in an age where assembly involved wire-wrap.
It was meant to be portable and MAINTAINABLE rather than 'no maintenance.'

Which is how you got monoliths. Because it *is* maintainable and extendable, easily.

@inthehands even if you know the stack the old software is on, and even if the original developers are on hand rewriting software on new tech is on of the hardest things you can do. With something this old, the code is the documentation and nobody really understands it.
Factor in critical timing and massive data and it makes it even more complicated.
Also there isn't so much COBOL material online so the gai won't even be able to guess well.
I can't see how this will even slightly work

@inthehands

this makes the various fiascos with the patent office software and IRS rewrites seem like massive perfection by comparison. and those took years, multiple attempts, insane amounts of dollars.

of course, the problem here is that they *WANT* to completely blow all this up and don't care about the suffering and possible deaths of actual human beings this could cause.

@inthehands oooh, this sounds interesting even without the added tension of a critical agency collapse

@inthehands the actual, spoken-out-loud plan is for the "old" system to collapse.

"Earlier this month, Yarvin cheered on the idea of hacking existing infrastructure “to operate in an unusual way that its designers, its previous operators, or both, did not expect,” and complimented DOGE for the way it has hacked into existing bureaucracies. The key performance indicator of DOGE, he wrote, “is its ability to take power from the libs, then keep it.”

https://heathercoxrichardson.substack.com/p/march-27-2025

@inthehands Just please, for the love of all that is good and holy, let there be one adult somewhere in the mix who makes sure the existing code is backed up safely.

Because you're 100% right, this is the stupidest idea yet from a group of people whose footnote in history will read "Oh yeah, those dickheads."

@inthehands

https://learn.microsoft.com/en-us/legal/cognitive-services/openai/transparency-note?context=%2Fazure%2Fai-services%2Fopenai%2Fcontext%2Fcontext&tabs=text

TLDR from the Microsoft OpenAI Transparency Note (the only place they have to be truthful, but are still allowed to weasel their way out of any responsibility):

• Don't use genAI when users can create text or code without end. You can get harmful or offensive content, especially with longer text.
• Don't use genAI when you need up-to-date information, or accuracy, or facts.
• Don't use genAI in ways that could cause real physical or psychological harm, like diagnosing patients or prescribing medications.
• Don't use genAI in ways that could impact legal rights, opportunities, or access to essential services like healthcare, education, or employment.
• Avoid using genAI where errors could lead to serious harm. GenAI is build using biased and undesirable data, and will produce biased and undesirable data.
• Don't use genAI in industries like healthcare, finance, or law, where mistakes can have serious consequences.
• Limit genAI as a chatbot to specific, well-defined chats to reduce the risk of unintended or inappropriate responses.
• Think twice (then think again) before using genAI, because it will often produce unintended results and require very strong guardrails.
• Organizations need to consider legal and regulatory obligations when using genAI, as it is probably not suitable for the industry or scenario. Make sure genAI use complies with applicable laws, terms and codes of conduct (we won't take the blame if it goes wrong).

@teledyn @inthehands I'm pretty sure climate change did not get the memo it does not exist and neither the Sui Dynasty nor the Neo-Assyrians were actively trying to ignore a structural threat, did they?

@thias @inthehands

Seems to me it might. Climate was certainly a factor bringing down the Minoans and just about everyone else c.1170 bce, but that took decades to accomplish, although the whole terrorism thing and immigration would be common talk back then.

@inthehands that would be not just disastrous for social security itself but for the entire software industry. The damage could happen even if they fail to actually replace anything but simply say they did it. A bunch of managers and investors are suddenly going to have absolutely absurd beliefs about what's possible in development. So many people are so credulous towards Musk because they want to believe him.

@thomasjwebb @inthehands wouldn't the absolute failure of such a high profile project expose the absurdity of the current AI hype ? Burst that damn bubble ?

(I certainly hope that this bubble will burst, sooner rather than later actually, but not at such a cost.)

@TofuTheSquirrel @inthehands I've witnessed a lot of choose-your-own-reality narratives around Twitter after Musk's takeover so I wouldn't underestimate their ability to pretend away inconvenient facts when problems happen. It does seem though that the bubble's in the process of bursting, but not everyone's going to get the memo at once.

@inthehands
(its us. we lost)

@inthehands I've been saying it for a while now because, sadly, it keeps being relevant; to Musk, failure is an option.

@buherator
To be fair, they didn’t specify the number of years. Or the number of digits in the number of years.

@thomasjwebb @inthehands

i don't GAF about the software industry, if they failed to actually replace anything but just said they did, that would be fine, Social Security as is works very well

the problem is they're going to break what's working & leave those of us who have worked since age 16 and not yet even reached the full retirement age of 67 with NOTHING to live on when we can't work after we made 50 years of contributions

everything else is a distraction

@peachfront @thomasjwebb @inthehands

I firmly believe one of the main goals of the musk/trump admin is to steal as much money as they can from the govt/taxpayers.

Not content with not paying taxes themselves to begin with, they're going to fleece us all.

@inthehands

The goals of billionaires is not to rewrite government functions, but to destroy government to obliterate the nation.

And if you think about it, that makes what what they are doing fully consistent. They may fantasize that they can do government better as authoritarian nut jobs but every time they’ve ever tried their insane libertarian ideas it’s ended in failure.

@inthehands

And of course, when they obliterate everything like this will very quickly understand just what AI is for.

The project simply exist to legitimize annihilating people people’s lives, killing them, killing men, women, children, stepping stones to their absolute rule in this little fantasy nightmare they called the network state.

@munin

Sing it out, and sing it loud... "If it works, and it ain't broken, don't 'fix' it" 😂

@inthehands @alice

@q

that is an -amazing- domain name lol

@rootwyrm @munin @inthehands @alice

Maybe COBOL's biggest problem is just that the developer base is shrinking, since people keep running after the next more sexy lang/framework?

Let's make COBOL sexy again(?) 😄

@henrik @munin @inthehands @alice man, don't get me started on Shiny Object Syndrome and fucking Silly-Con Valley Cargo Culting.

But hey, if we force everyone to start over again every 3 years, we can justify keeping their pay flat or even decreasing it. >:(

@rootwyrm

Listing: "Wanted: 5+ years experience in shiny new language that's been around 3 years."

Interviewer: "Well, your resume is impressive, but you don't meet all the requirements, so we might not be able to meet your asking price."

@henrik @munin @inthehands

@munin @inthehands @alice also, generative ai produces more insecure code. These kids are being stupid and are supposedly the geniuses of Gen Z. sigh...

@munin @inthehands @alice As someone who has programmed in COBOL my entire (nearly) 3 decade career, I totally agree. My clients are banks. And they usually say that my system is the most stable they know of in the bank.

@inthehands

As a retired person who collects SS payments, this is – to put it mildly – very concerning. As a computer programmer who (a very long time ago) helped maintain a large COBOL codebase?

Shocked. Gobsmacked at the sheer arrogance required to even say you are going to do something like that. This is not audacity; this is the captain of the Titanic seeing an iceberg and ordering, "Full speed ahead!"

FWIW? I have some other resources to cover short term. I know others on SS who do not.

@inthehands If I was looking for a way to destroy social security systems, doing it with AI would be up there as the most wasteful method.

@inthehands At the end of the article comes this quote:
“The leaders need to understand that they’re dealing with a house of cards or Jenga. If they start pulling pieces out, which they’ve already stated they’re doing, things can break.”
Broski, that is exactly what they want to do, break the system.

@inthehands

No sane management would EVER agree to that process:

It's a HELLA lotta work.

Translating from one language to another provides NO benefits to the end-user.

Look -- I read the Wired article, too. It convinced me that the Social Security system absolutely needs good engineers to update it -- anything both critical and fragile needs better engineering. But the way to do is is exceedingly BORING (maintaining Blue-green systems that demonstrate that they produce exactly the same output) and painstaking.

There used to be a team in the Federal government that did exactly this kind of stuff... oh, yeah -- https://federalnewsnetwork.com/reorganization/2025/03/after-rocky-history-gsa-shuts-down-18f-office/

@inthehands Move fast and break (other people’s) things.

@inthehands Those millions of lines of COBOL code contain a huge amount of policy (both organizational for the service itself, and for implementatrion, for instance a flag equal to "Z" may mean that person has received benefits in past but no longer needs them even if still alive.
The kids writing java code wouldn't have this institutional kowledge and be clueless on all the possible edge cases the system needs to handle.

@henrik @rootwyrm @munin @inthehands @alice

it should not because is readable by judges accountants lawyers et cetera ...

but thats what they hate ... ofuscate to steal ... thats the plan ... insecurity by obscurity ...

@inthehands it’s stupid if you want social security to function. If your goal is to break it so severely that privatizing it looks appealing—which has been the prerogative of these shitheads for 40 years—then this idea is exactly in line with their goals

@thedansimonson @inthehands this!!

Like, I don't understand why anybody is still trying to pretend that we're supposed to be having a "competence" argument.

a) they won the vote. They are implementing policy at astonishing rates. They are enormously competent at achieving their goals.

b) they are not even running on competence. They are running on grievances. arguing that they're doing a bad job is not how you will win people over.

c) they are very clear about their end goals. People need to stop reacting to inverted dog whistles. Stop arguing on their terms. They aren't trying to improve social security. They just say so to keep just enough naive morons quiet.

If you catch an axe murderer in the act, you don't start arguing that well akshually, he's holding a chisel. You call him a murderer.

@teledyn @inthehands Republicans have been hacking away at the foundations of our country for 40 years.

@mikemccaffrey @inthehands

https://youtu.be/Gr7T07WfIhM?si=jYBl5WChH-0BDOZJ

@TofuTheSquirrel @thomasjwebb @inthehands

I'm patiently waiting for Americans to wake up and throw these mad men in prison or smthg so maybe that massive fk up will be the spark they need?

It sucks for anyone involved ofc but I feel the sooner they rip off the bandage, the better :S

@Beldarak It's only Trump supporters who cannot see these problems you describe.

Sadly, many of them are totally ignorant about politics (knowing far less about it than many non-Americans) and would have to be personally affected by Musk or Trump in order to know about it.

@inthehands US Citizens will die because of this, and the US Government will not care.

@inthehands One thing I've learned about big, complicated systems built over a period of many years is that they are more often than not that way for a reason, and you'd better understand the reason before you mess with them.

@mattblaze Chesterton’s Fence, at the ripe old age of old I just heard of this recently. @inthehands

@Nonya_Bidniss @inthehands Exactly

@mattblaze @inthehands Matt, I have every confidence that the intellectual rigor and keen analytical skills necessary to a successful outcome is exactly what we’ll get from Big Balls and his esteemed colleagues.

@20002ist @inthehands Only the best people.

@inthehands @mattblaze
What exactly do they think will happen if their new system neglects to pay out expected benefits to millions of people?

I need to research guillotine manufacturers to invest in this rapid growth market.

@tsturm @inthehands They will cite the failure as evidence that the government is incompetent, and that social security should be privatized. They will do this without irony or even a hint of a smirk.

@inthehands But they reduced the public support staff *before* they refactor the entire system. Everyone might want to protect themselves by downloading any statement or reporting to use should things not line up on the other side

@inthehands

You might share attached. It took me five minutes to get my info. They’re bound to fuck it up..

@mattsheffield

What I meant is people not in his camp should do something (and at this point I'm not excluding violence, sadly).

I know it's easier said than done of course, and maybe it's just a lack of covering from press in my country, but I'm surprised I don't hear about factories on strike or tons of people in the streets.

Ofc it's hard to say what I'd do in the same situation but I like to think I would act against litteral nazis :S

@mattsheffield

To put things in context, I live in Belgium. I've been conditioned (pasively, just by our history) all my life to absolutely loathe nazis.

Every time someone brings a "what would you do with a time machine?" question most people here will eventually say "kill Hitler" in their propositions.

I'm a pacifist, anti-war, I used violence maybe once in my life (to get out of a situation). But nazis? That's where people draw the line here and agree: beat the shit out of them.

@mattsheffield

I visited the Dachau concentration camp when I was a teen, fuck this shit, don't let that happen in your country.

@Beldarak Unfortunately, television and social media are so entertaining for most Americans that they have no idea that they could have something better, and that they deserve it.

I am working with some people to change that though, but we need financial support. It's tough having to cold-start democracy in a country where it's been neglected.

@Nonya_Bidniss @mattblaze @inthehands

I've got lots of fences at the barely functional farm.
Some are just a couple of posts to mark specialty trees, some of which would cost a few hundred dollars to replace.
Another is just a weird arrangement of logs to remind people that inside that area there are a bunch of holes that copperheads like breeding in.
There's a series of stakes marking the official wetland boundary, if we plant crops below those we lose any chance of USDA grants and could possibly go to jail if there was an EPA enforcing laws.
And there's the 3D deer fence with 20kV as the outer layer.

If you don't know why the fences are there and decide to ignore them, you are likely having a Find Out moment.

@inthehands nobody will object until it actually DOES effect them.

of course, by then its too late, hehe.

Head-in-the-sand leads to whats gonna happen here (a bunch of buried heads)
I need more popcorn...

@inthehands holy wow. I always used to say that it was the hallmark of a senior engineer, that they had learned to resist the urge to rewrite it all. I guess Elon isn't exactly a senior engineer, so that checks out.

@inthehands And they gonna test in production. Including human sacrifices.

@mattblaze

Should we call this “Chesterton’s Program” or “Blaze’s Fence”?

@inthehands
I don't think succeeding is the point.
I think the point is creating an acceptable scenario in which to break it.
"Acceptable" meaning people don't riot.