One consistent thing about Musk and DOGE is the continuously promised use of Large Language model AI technologies to do…something.
The most recent case is Musk’s promise to use AI to analyse the five bullet responses to his “What did you do last week?” question…which government workers may or may not have been told to respond to or not respond to. Maybe. Yes. No. Maybe.
(BTW, there’s a lawsuit for that.)
Remember my reference to GIGO? Garbage In, Garbage Out? When you’re automatically collecting and analyzing data from many different groups with jobs that have no commonality working with hundreds of agencies enforcing rules, regulations, and requirements—all of which come from laws built on top of other laws. which may contradict each other in any number infinitesimal ways…well, the resulting mess will be pure GIGO. Particularly when answering may or may not happen individually, by employee, or by department. Especially when directions in how to respond come via X-witter twip.
The same GIGO applies to every stated use of AI that Musk and DOGE claim will root out waste, Fraud, and abuse (WAF). The same applies to all their claims about ‘modernizing’ the government computer systems.
The idea of being able to reduce decades-old government systems to something that can either be understood or redefined automatically is equivalent to reducing SpaceX software systems to “Rocket go up. Rocket come down.”
I took a deep dive into the systems used in our government agencies. There’s a surprising amount of openly available data about each.
For instance, we know about COBOL used with government systems, but are you aware that the core element of the Treasury fiscal system is down-to-the-metal Assembly code?
More modern systems utilize Java, and many are created using C and C++. However, in 2025 government completed a two year study about what language it should be using because C and C++ are inherently insecure.
Their answer? Rust.
After more than two decades of grappling with memory safety issues in C and C++, the software engineering community has reached a consensus. It’s not enough to rely on bug-finding tools.
The preferred approach is to use “safe” programming languages that can reject unsafe programs at compile time, thereby preventing the emergence of memory safety issues.
The TRACTOR program aims to automate the translation of legacy C code to Rust.
More here.
I find it interesting that SpaceX primarily uses C and C++. However, it also uses a smattering of Python, Node.js, and FORTRAN so maybe there’s hope the company’s programmers can pick up Rust.
Speaking of FORTRAN … ah, the king of complex computations. Unlike COBOL, FORTRAN will never die…it will just kind of linger for a long, long time. Not as long as C, but long.
FORTRAN was big with the military at one time, along with other languages, many of them proprietary. In the early 1980s, the military created their own language, Ada (not to be confused with Air Defense Artillery), a cousin of Pascal and Module 2. I don’t know Ada, but I do know Module-2—none of which should be confused with the Module 2 of the coursework for ADA, or the Americans for Disability Act.
(The government is really big on acronyms. So much so, it keeps a list. Well, it keeps many lists. But no worries, there’s a guide.)
FORTRAN is still utilized in government systems, such as those used by NOAA with a callable interface to their MADIS API. In fact, did you know the Commerce department supports 53,752 searchable data sets in their Commerce Data Hub? Who knows the backend language used for each, but they provide either a REST or SOAP interface. And JSON!
NOAA has 46,861 data sets, alone. I wonder what kind of private enterprise would be willing to provide the same?
And these are data sets that DOGE and Trump’s admin haven’t destroyed, yet. Which really peeves me. As a software engineer who primarily worked with data, destroying any data is a sacrilege. It’s also a massive waste of government funding. We, the citizens of this country, paid for the data and documents. We, the citizens of this country that actually pad taxes, paid for the data and documents.
What kind of people would destroy data? Good data, useful data? Not the kind of people I’ve worked with in almost 40 years of software engineering.
Anyway, the point is made: the federal government systems are incredibly large, complex, and most don’t even share a common programming language. The DOGE kids may be wiz bang at making rockets go up, and rockets come down, but these Big Egos have never met Big Iron. They’ve never had to deal with legacy systems like those of the federal government. They’ve likely never had to deal with legacy systems, period.
More importantly, they have no respect for anything outside their experience. They have no respect for the complexity of the systems, the importance of them operating without fail, and all that wondrous data.
They’re lost and it shows by the crude chopping block approach to save a few bucks, and that will cost us double in the long run. And it shows by the number of errors that continue to show up on their cute little scoreboard.
They are outclassed.
I know big systems. I worked on one once for Boeing Military. It was in FORTRAN. I’ve also worked on other big systems, but most of them were in Java or C or C++. I’m fond of Node.js, but I’ve not worked with a big system with it. I’ve only incidentally seen systems created in COBOL. But…I know big systems.
And you couldn’t pay me to walk into any one of the IT departments for these government agencies with a little Starlink-connected laptop and cyber-insecure cloud app, and pretend to be some kind of über intellect who is going to automagically change things for the better. And I know there is more than one old code warrier like me out there nodding their heads in agreement.
Note: not making fun of SpaceX programmers. They do excellent work. But programming new software for a spaceship is not the same as understanding decades-old and complex federal systems created in dozens of programming languages, each of which has to follow law, not logic. Not the same thing.
