by SeenNotHeard on 2/16/25, 11:56 PM with 499 comments
by Aloisius on 2/17/25, 3:24 AM
https://www.openicpsr.org/openicpsr/project/207202/
I have a hard time believing the DB2 systems would convert it to days/seconds/whatever since 1875. It's not impossible, but I think whoever came up with the 1875 thing was simply wrong.
That's not to say there are 150 year olds collecting social security either. Dates of birth are sometimes missing or entered wrong and sometimes death records don't get entered. It's also clear DOGE didn't understand that social security numbers can't be used as a unique identifiers (nor why it's unnecessary) which can lead to all sorts of issues when processing.
Edit: It also seems the SSA presumes anyone over 115 is has died and halts payments which makes it even more unlikely there are 150 year old beneficiaries: https://secure.ssa.gov/poms.nsf/lnx/0202602578
by bagels on 2/17/25, 9:24 PM
They use multiple techniques and data sources to determine who to send benefits to.
This is not news to the SSA.
https://oig.ssa.gov/assets/uploads/a-06-21-51022.pdf
"AGENCY COMMENTS SSA disagreed with our recommendations. Agency officials stated that most of the records discussed in the report involve numberholders who do not currently receive SSA payment"
So, they can do better, but sure, they are sending some relatively small number of checks out to dead people. That doesn't mean Musk needs to lie about the the program as an excuse to cut the whole thing, which is actually what we see playing out.
by cwbriscoe on 2/17/25, 1:52 AM
Pre-Y2k a lot of dates were created with:
ACCEPT WS-DATE FROM DATE.
The above WS-DATE was in YYMMDD format, which is why there was a Y2K issue and needed to be resolved with windowing code. However, windowing code wouldn't work for somebody that was over 100 years old...
Doing a little research there is also (which I have never used since we just use DB2 and date parameter input files for CCYYMMDD dates):
ACCEPT WS-CENTURY-DATE FROM CENTURY-DATE.
This date is in CCYYMMDD format. According to google, the epoch for this date is January 1st, 1601.
by hypeatei on 2/17/25, 1:10 AM
by Devasta on 2/17/25, 9:32 AM
Prove Musk wrong on this and he'll just go about his day as normal; 20 minutes later there'll be someone tweeting an unhinged screed about how the US government is spending 10 trillion this year changing the name of the Department of Homeland Security to the Department of Homeland Inclusivity to which he'll quote tweet "Interesting" and then he'll set his little band of freaks to cause mayhem somewhere else.
by spullara on 2/17/25, 7:14 AM
https://x.com/elonmusk/status/1891350795452654076
...
100-109 4,734,407
110-119 3,627,007
120-129 3,472,849
130-139 3,936,311
140-149 3,542,044
150-159 1,345,083
160-169 121,807
170-179 6,087
180-189 695
190-199 448
200-209 879
210-219 866
220-229 1,039
240-249 1
360-369 1by totallynothoney on 2/17/25, 4:04 AM
[0]. https://en.wikipedia.org/wiki/American_Civil_War_widows_who_... [1]. https://en.wikipedia.org/wiki/Ida_May_Fuller
by layer8 on 2/17/25, 2:10 AM
by jiggawatts on 2/17/25, 3:49 AM
----
There was a time when data structures were made to fit purpose, not compilers. Having a look at the subject, shows clearly the constrains for valid dates:
Social Security was introduced in 1935.
To be eligible for benefits one had to
pay in at least 40 quarters, that's 10 years
be at least 65 years old
So add another 5 years for wiggle room and we end at a nice round 1875 as earliest year for any birthday to be recorded.
A perfect rational base for a date entry, isn't it?
----
The "COBOL doesn't work like that" comments are missing the forest for the trees: This is a very old system with bespoke coding to match legislation, not legislation to match compiler default behaviour.
Fundamentally, unless a government employee that has worked directly on this codebase speaks up, we're all just guessing.
by roshin on 2/17/25, 5:25 AM
by noobermin on 2/17/25, 5:59 AM
by bobnick on 2/17/25, 2:15 PM
by paulsutter on 2/17/25, 5:32 AM
(hint: there is a smooth trend of people of all ages up to 199, so the 1875 thing was pure misdirection)
https://x.com/elonmusk/status/1891350795452654076?s=46&t=NN3...
by ZeroGravitas on 2/17/25, 9:48 AM
https://oig.ssa.gov/assets/uploads/a-06-21-51022.pdf
Numberholders Age 100 or Older Who Did Not Have Death Information on the Numident
> The attached final report presents the results of the Office of Audit’s review. The objective was to determine whether the Social Security Administration had effective controls to annotate death information on the Numident records of numberholders who exceeded maximum reasonable life expectancies. Please provide within 60 days a corrective action plan that addresses each recommendation. If you wish to discuss the final report, please call me or have your staff contact Michelle L. Anderson, Assistant Inspector General for Audit
They decided not to do anything about it (e.g. add a "presumed dead" field) because they thought it would be a waste of money!
> In response to our 2015 report, SSA considered multiple options, including adding presumed death information to these Numident records. SSA ultimately decided not to proceed because the “. . . options would be costly to implement, would be of little benefit to the agency, would largely duplicate information already available to data exchange consumers and would create cost for the states and other data exchange partners.”16 SSA also believed a regulation would be required to allow it to add death information to these records, and adding presumed death information to the Numident would increase the risk of inadvertent release of living individuals’ personal information in the DMF.
Submitted here in case anyone wants to discuss the SHOCKINGLY boring REVELATIONS contained within:
by Sniffnoy on 2/17/25, 3:49 AM
by hans_castorp on 2/17/25, 6:11 AM
However, the records were never stored for more than 4 years, so this was never a problem.
by blindriver on 2/17/25, 1:03 AM
With so much manipulated information, AI-generated content, and straight up lying, I really can't tell what's real and fake anymore.
I distinctly remember finally not being able to tell the difference between fake and real info during the Allen Texas shopping mall shooting. I went on Twitter to get more info and I couldn't tell what was real and what was fake for the first time because everything was so convincing. That feels like ages ago now because things are so much more sophisticated.
by blame-troi on 2/17/25, 1:15 PM
This isn't a COBOL issue (if it's even an issue at all), it's a data design issue. As many have pointed out, there are reasons for this possible origin date.
by jmclnx on 2/17/25, 1:32 AM
Also I never head of this default.
by SandyAndyPerth on 2/26/25, 6:38 AM
Nobody in this HN thread has used the word "sentinel" - see another HN about the concept https://news.ycombinator.com/item?id=36195425
People got hung up on: - "COBOL defaults to..." rather than "banking practices are..." - epoch start dates - many pointing out COBOL didn't use epochs or counts, just much-damned YYDDD or YYMMDD actual strings.
Also, Elon loves to stir with partial misinfo hence his tweet https://x.com/elonmusk/status/1891350795452654076 with the breakdowns by age bracket. "Death set to FALSE" means "Death date not known" but that's not clickbaity enough.
That long tail looks awfully like data entered from historical records lacking death dates - there have been a few discussions of the cost of finding death dates and the decision to avoid spending $millions on it, as this is not data used to make payments.
You would expect, in a system that's pulling data from many sources, to see historical jumps in data cleanup like this. Imagine a few large states finally get around to digital records of deaths, so their data is easily aggregated - you get a sudden flushing of people who would previously have been left on the list. However, this will only apply from a certain age onwards as those sources in turn don't have the time/budget/interest to digitise really old records.
by Calzifer on 2/19/25, 3:26 PM
Assuming the reference date is correct and unknown birth date is stored as timestamp 0. Since we are before 2025-05-20 at the moment the reported age has to be 149 years, not 150. What I'm missing? Would be very unusual to round up age.
by n0denine on 2/20/25, 6:47 PM
by throw0101d on 2/17/25, 1:48 PM
* https://lucid.co/techblog/2020/11/13/why-cobol-isnt-the-prob...
by MrCOBOL on 2/17/25, 3:17 PM
by hbarka on 2/17/25, 1:48 AM
by dashundchen on 2/17/25, 4:30 AM
Musk, Trump and the admin have already been pushing so many outright lies via their propaganda channels. Despite being deunked the lies are repeated nonstop.
The lie that USAID spent $50 million on condoms in Gaza (the money was for running hospitals) for example.
Or the lie about funding an opera about a transgender woman (the money was for a university in Columbia, unrelated to a performance put on at the school).
They are spewing lies left and right. Lies go twice around the block before the truth has put its pants on.
Why should we believe him on this?
by Modulius on 2/18/25, 6:53 PM
https://x.com/KariLake/status/1891841704703013067
Look at the comments. They know that COBOL defaults to 1875, point is that propaganda pundits look for any reason to spew toxic misinformation and rile 99.999% of uniformed sheeple that voted for orange felon.
by jedwards1211 on 2/19/25, 7:02 AM
by UltraSane on 2/17/25, 8:43 AM
by jamesrom on 2/17/25, 6:54 AM
He qualified this claim as a “cursory examination”. It’s clearly a comment about the quality of the data and systems. That this is the kind of thing that would be prone to fraud.
Before you hit downvote, please provide evidence that you didn’t hallucinate Musk’s claims here.
by NoPicklez on 2/17/25, 3:24 AM
There’s been no formal inquiry, there was no mention of any of the checks and balances that may have been occurring, there’s been no nuance to the argument from him at all. Okay maybe there are people that look to be 150 years old, is there a reason why? Were they actually being paid the social security or was there a legitimate exception? Maybe people were but it was so few and far between and was an internal controls issue which all governments and companies have globally or was none of it true and your team simply assumed they were being paid.
After reading an article where Gov security experts were worried because they were having to give Elon’s team access to Putty and SQL tools. It seems like people are going through this data and making inferences that may not be completely true or vetted.
by laurent_du on 2/17/25, 11:30 AM
Yet everyone's conclusion is, as usual, Musk bad. Nobody cares about left-wing fakes news if they are useful to the agenda being pushed forward.
by msie on 2/17/25, 10:43 PM
by cjbgkagh on 2/17/25, 3:45 PM
Someone would have to be really incompetent to find a spike at precisely 150 years old and not investigate it further. Elon tweeted ~ 10 hours ago the age breakdown and there does not appear to be a spike at 150 so if that information is correct then this is no longer evidence of incompetence.
by ggm on 2/17/25, 1:11 AM