Great piece. I am very distressed by the running around and trying to break everything for the sake of “auditing” and/or misappropriating the power of the purse from Congress. Even more I’m just distressed by how he and Trump are controlling the narrative about these things being “good” when without a doubt not all of it, maybe not even most of it, will be.
It's a misnomer to call any of this an audit: that would involve forensic accountants, at a minimum. Importantly, an audit is used to gather information about what's what and to then provide guidance on areas where changes can be made. That takes time: a good example being Clinton's efforts to reduce in the federal workforce took 6 months just to gather the info and identify who/what could be cut.
The system is not worth saving. Its corrupt. Your inability to imagine how it can be done is not in and of itself proof it can’t or shouldn’t be replaced. (Yeah, I programmed Fortran and Cobal on punch cards too. Not the bragging point you think it is.)
I won't pretend to know very much about the subject. I was forced, in my late 40s, to take a few continuing education classes to learn a little about file types, BASIC programming and Microsoft systems because my company was getting away from IBM mainframes including the one used at my location. I had to learn the vocabulary because our local IT department reported to me during the changes. I do not recall any real problems with the process or with coping with the Y2K panic 12 years later.
It seems to me that getting very bright people involved in the legacy systems is a good move. They might learn quickly and find ways to modernize the systems. Why couldn't the data be read into new formats that would allow legacy obsolete data to be archived in an orderly manner? One recent comment on this general subject was that only 13 people over 112 years old were drawing social security checks. I don't understand how this is known if the legacy systems are such a deep mystery but there could be a cutoff point at the birthdate of the oldest of those thirteen. A modernized system should be able to generate reports that require follow up on recipients over a certain age. Maybe that would help expose cheaters who are keeping mom and grandma in the freezer and living off their social security payments.
You don't typically update a running production system directly when you modernize stuff like this. You stand up development and staging servers and make your changes there, text exhaustively, and once everything checks out, you switch over from the legacy system to the new one.
NONE OF THAT is happening from the reports coming out of the agency. You have youngsters with HUGE Dunning-Kruger complexes futzing around with running systems that they haven't put in the time to understand.
I actually know that much from experience. My understanding is that they are extracting and tabulating data in an effort to understand what they are dealing with. They are getting plenty help from DOGE critics. No one has said they are changing the system or even ready to change it.
There's reporting that they're opening up back doors into live systems, which does require modifications to the systems (and we can discuss the security implications of this elsewhere), as opposed to asking for data and file dumps that they can analyze on non-mission-critical systems.
The way to look at the SSA info that is being referred to is that it is in fact multiple databases/systems that are involved: one which appears to be providing somewhat basic info on SSNs and age of the associated person, and others that involve tracking payments. That's how we can know both how many active account numbers there are for a particular age, and how many of those receive payments.
We can then say that one of those systems - the one involving active account numbers - is broken in a particular way that ultimately has been found to not cause a problem in other systems.
So this brings us back to the question of whether it's worth "fixing the problem", or whether we live with it because the cost outweighs the benefit (as SSA has stated in their responses to the OIG audit recommendations).
As long as there is practically free tech help, I would say fix it. All of you more experienced experts can chime in as you see fit. I would but I am not an expert. And I'm sure you have jobs. Probably explains the age of the DOGE staff.
We have. Chris provided a decent high-level of what the process looks like: know how everything connects, test the desired changes, iterate until it's ready for push to production.
Nope. But I do know how to write software and deal with legacy computer systems. Let me know when rebuilding a big block GM V8 is a required skill for that.
Just reading this today because I confess sometimes , though I've followed Racket News since you started, I'm just tired of the Trump bashing. But, I'm glad I opened it. While it's way over my head, I get what you are saying. Great read. Thank you very much for the explanation.
Semi-related: the administration is now moving to cut the Board of Immigration Appeals from 28 judges to 15. This comes after having fired 20 "probationary" judges.
This will lead to further delays and longer lines, making the system even worse efficiency-wise than it already was - with 3.7mil cases in the backlog.
Great piece. I am very distressed by the running around and trying to break everything for the sake of “auditing” and/or misappropriating the power of the purse from Congress. Even more I’m just distressed by how he and Trump are controlling the narrative about these things being “good” when without a doubt not all of it, maybe not even most of it, will be.
It's a misnomer to call any of this an audit: that would involve forensic accountants, at a minimum. Importantly, an audit is used to gather information about what's what and to then provide guidance on areas where changes can be made. That takes time: a good example being Clinton's efforts to reduce in the federal workforce took 6 months just to gather the info and identify who/what could be cut.
The system is not worth saving. Its corrupt. Your inability to imagine how it can be done is not in and of itself proof it can’t or shouldn’t be replaced. (Yeah, I programmed Fortran and Cobal on punch cards too. Not the bragging point you think it is.)
I won't pretend to know very much about the subject. I was forced, in my late 40s, to take a few continuing education classes to learn a little about file types, BASIC programming and Microsoft systems because my company was getting away from IBM mainframes including the one used at my location. I had to learn the vocabulary because our local IT department reported to me during the changes. I do not recall any real problems with the process or with coping with the Y2K panic 12 years later.
It seems to me that getting very bright people involved in the legacy systems is a good move. They might learn quickly and find ways to modernize the systems. Why couldn't the data be read into new formats that would allow legacy obsolete data to be archived in an orderly manner? One recent comment on this general subject was that only 13 people over 112 years old were drawing social security checks. I don't understand how this is known if the legacy systems are such a deep mystery but there could be a cutoff point at the birthdate of the oldest of those thirteen. A modernized system should be able to generate reports that require follow up on recipients over a certain age. Maybe that would help expose cheaters who are keeping mom and grandma in the freezer and living off their social security payments.
You don't typically update a running production system directly when you modernize stuff like this. You stand up development and staging servers and make your changes there, text exhaustively, and once everything checks out, you switch over from the legacy system to the new one.
NONE OF THAT is happening from the reports coming out of the agency. You have youngsters with HUGE Dunning-Kruger complexes futzing around with running systems that they haven't put in the time to understand.
Process is important.
Exactly
I actually know that much from experience. My understanding is that they are extracting and tabulating data in an effort to understand what they are dealing with. They are getting plenty help from DOGE critics. No one has said they are changing the system or even ready to change it.
There's reporting that they're opening up back doors into live systems, which does require modifications to the systems (and we can discuss the security implications of this elsewhere), as opposed to asking for data and file dumps that they can analyze on non-mission-critical systems.
https://www.theatlantic.com/technology/archive/2025/02/doge-god-mode-access/681719/?gift=Eup24Na3uAgSVBW80exvj_--_aVv-MSeUQSghZZfKr4&utm_source=copy-link&utm_medium=social&utm_campaign=share
A lot of speculation and anonymous sources.
They don’t know what they’re extracting. GIGO
So far.
And yet they're stating things definitively.
They are reporting on the data extracted. What it means is anybody's guess.
The way to look at the SSA info that is being referred to is that it is in fact multiple databases/systems that are involved: one which appears to be providing somewhat basic info on SSNs and age of the associated person, and others that involve tracking payments. That's how we can know both how many active account numbers there are for a particular age, and how many of those receive payments.
We can then say that one of those systems - the one involving active account numbers - is broken in a particular way that ultimately has been found to not cause a problem in other systems.
So this brings us back to the question of whether it's worth "fixing the problem", or whether we live with it because the cost outweighs the benefit (as SSA has stated in their responses to the OIG audit recommendations).
As long as there is practically free tech help, I would say fix it. All of you more experienced experts can chime in as you see fit. I would but I am not an expert. And I'm sure you have jobs. Probably explains the age of the DOGE staff.
We have. Chris provided a decent high-level of what the process looks like: know how everything connects, test the desired changes, iterate until it's ready for push to production.
Since you don’t know how to rebuild a 454, I’m supposed to accept your take on reducing theft and waste? Yep, Fetterman is your guy. Lol
Nope. But I do know how to write software and deal with legacy computer systems. Let me know when rebuilding a big block GM V8 is a required skill for that.
Just reading this today because I confess sometimes , though I've followed Racket News since you started, I'm just tired of the Trump bashing. But, I'm glad I opened it. While it's way over my head, I get what you are saying. Great read. Thank you very much for the explanation.
Semi-related: the administration is now moving to cut the Board of Immigration Appeals from 28 judges to 15. This comes after having fired 20 "probationary" judges.
This will lead to further delays and longer lines, making the system even worse efficiency-wise than it already was - with 3.7mil cases in the backlog.
Immigration judges are union employees. They have no use for productivity standards. Every case is treated as a new learning experience.
It's our immigration system: it only works if you can process requests in a timely manner.