DOGE plans to rebuild the Social Security Administration COBAL code base in months. Doing so is a danger to benefits and could cause a system collapse.

Social Security systems contain tens of millions of lines of code written in COBOL, an archaic programming language. Safely rewriting that code would take years—DOGE wants it done in months.

65 million people in the US currently receive Social Security benefits. They're considering Java as a replacement.

The project is being organized by Elon Musk lieutenant Steve Davis, multiple sources who were not given permission to talk to the media tell WIRED, and aims to migrate all SSA systems off COBOL, one of the first common business-oriented programming languages, and onto a more modern replacement like Java within a scheduled tight timeframe of a few months.

SSA’s core “logic” is also written largely in COBOL. This is the code that issues social security numbers, manages payments, and even calculates the total amount beneficiaries should receive for different services, a former senior SSA technologist who worked in the office of the chief information officer says. Even minor changes could result in cascading failures across programs.

46 comments

Converting a system from COBOL to a modern system is not something you do in a few months. A quote from Bill Hinshaw of COBOL Cowboys:
“I just got through a conversion (for a system) to go from COBOL to Java, It’s taken them four years, and they’re still not done.”
I’m sure whatever system Bill was referring to was of a much smaller scale than the social security system. For another example:
Commonwealth Bank of Australia, for instance, replaced its core banking platform in 2012 with the help of Accenture and software company SAP SE. The job ultimately took five years and cost more than 1 billion Australian dollars ($749.9 million).
And, of course, let us not forget the massive fuck up of TSB’s migration off COBOL. That migration was done for no more than three years, and the rollout, was, well:
U.K. bank, TSB, was forced to migrate from a COBOL-based system in 2018 due to a buyout. It didn't go well. Because the bank was unable to trade for days, the cost of the migration ended up being 330 million pounds. That was in addition to the budgeted cost for the engineering work for the actual migration. TSB also lost 49.1 million pounds from financial fraud while its systems were melting down.
Customer compensation topped 125 million pounds, and the bank had to spend 122 million pounds hiring new staff to deal with the 204,000 customer complaint cases. The chief executive resigned and the company is still mopping up the damage two years after the event.
TSB also ended up getting fined £48.7 million by UK regulators for their mishandling of the transfer. Mind you, TSB is not a mega-bank, they’re a modest sized one, and they had Accenture doing it who had way more experience with this sort of thing than Musk and his skibidi crew. If Musk follows through on this, we’re talking massive numbers of people unable to access their payments, economic fallout most likely triggering a recession, god knows how many people having their personal data compromised, and subsequently a level of fraud that will make whatever minor fraud is going now look like a teenager shoplifting a pack of gum.
- The difference here is that those banks presumably at least wanted their systems to keep working after being migrated off COBOL. For the
  crew, making the system completely broken and unusable is a bonus objective.
  
  One of Robert Moses’ key ways to build his empire and power was, get stakes in the ground as soon as possible. Doesn’t matter if the plans are finished or if not all the funding is there, just get it started and then the politicos will be pressured into finishing to save face of having a failed project.
  Musk gets this speed run out the door, doesn’t matter if it’s a boondoggle on launch, as long as it’s his team that has the spaghetti code in their pocket, that creates the impetus to “let him finish the job” and then one of his corporate fiefdoms will become the one that manages the SSA system in perpetuity.
  
  I could definitely see this happening, if somehow in four years dema get elected they'll probably be too cowardly to stop any of these plans once they're set in motion out of fear that they'd be wasting a sunk cost.
- Why would you even want to move from COBOL anyway? Why not just train people to learn it despite it's age?
  
  This is a bit apples to oranges, but a few years back the DoD transferred their logistics system from COBOL to Java for the following given reasons:
  A major component of the system is 54 years old, written in COBOL, and provides retail-level business logic. The component runs on mainframes that have proven to be extremely difficult to change and manage, and the DoD needed to modernize the component to drive down operating costs and move to an open platform, while retaining all functionality...
  After 54 years of operations, maintenance, and extensions, the component’s code had become poorly documented. The technical design of the existing system, which was needed to support the modernization effort, had to be derived from the existing system and code.
  There’s a sensibility to that; logistics is an area that’s more likely to be interfacing with more modern systems than payments, and if the documentation has become an indecipherable mess, that’s a problem waiting to happen if something goes wrong. Oh, and worth noting that relatively small system took nearly three years to transfer to Java.
  However, as I noted, I don’t think any of that is the motivation here, it’s all about placing the SSA system under Musk’s umbrella.
  
  Couldn't the part running on a mainframe be virtualized and run on basically anything at this point?
  
  It could, but this only solves the problem of hardware availability and throughput. It does noting with regards to maintenance, system architecture, or adapting the implementation to new requirements imposed either by legislation, policy, or changing technology (or mission creep, in the case of the DoD). You're still stuck with a hundred or so million lines of code which nobody understands.

46 comments