On DOGE and government efficiency: bad software, bad regulations, underfunded commitments and more

I spent six years as a mid-level (regional) federal bureaucrat in VHA. I'm also a tech and economics geek with 20 years experience at a very large publicly traded corporation and through our children I have extensive experience as a consumer of federal government services. Lastly I don't have to worry about offending employers or colleagues.

All of which is to say that if you want to think about President Musk's government efficiency push these thoughts may be helpful.

1. The Federal government hires a lot of people to work around both software limitations and a rats nest of Congressional mandates and regulations that may have been well intended but are now obsolete and harmful. The mandates and regulations are only added, never removed.
2. Much of the federal government is outsourced to huge consulting firms that benefit from both less awful internal software and freedom from many federal mandates and regulations (famously including many laws constraining law enforcement). They take a large cut for themselves so their net efficiency effect might be positive or negative. If government became more efficient then much of that outsourcing should be reversed -- but that would drastically reduce campaign donations and post-political employment for public servants.
3. Government software quality was pretty good in the 80s and into the 90s. I've seen some good work in past few years, often by a very quiet semi-volunteer SV cohort that was brought in by Obama, survived Trump I, and continued through Biden.
4. The Federal government lives by unfunded mandates and underfunded commitments. Inefficiency in many forms is a major way to reduce spending on those commitments. A more efficient government would employ far fewer people but would spend more money. The net effect might be negative.
5. If software quality were improved, even without removing dysfunctional internal mandates and regulations, large numbers of working class American would lose what are often low paying but secure high benefit jobs. Even as the overall economy has far less to offer the non-academic class than it did even 20 years ago. Inefficient government employment is a model for what we will need to do for most Americans as our AIs develop.
6. The main reason federal software quality is poor is because federal software procurement follows rules set up in the 1980s to reform defense department purchasing and because of mandates designed to support military veterans and historically disadvantaged populations. These rules are incompatible with creating and maintaining complex software.
7. If government switched to modern software platforms there would be a huge expenditure to upgrade IT infrastructure.

That's about all I have. I know there are others who can offer far more detail, but they will usually be unable to speak freely. If you are a journalist I suggest hunting down other retired mid-level federal bureaucrats.