Thanks for the perspective. I would like to note that pointing out earlier instances of bad behavior is important historical context, but doesn't make more recent bad behavior any less bad. The USA has some really dark stuff in its past. We should remember and do better.

One major category of concern is institutional erosion. Here are a few salient examples:

1. Mass IG firing (Jan 2025): Fired 17 inspectors general in one night without the legally required 30-day notice. Federal court ruled this "obviously" illegal. Several were investigating Musk's companies. (IGs are the internal watchdogs that catch waste, fraud, and abuse.)
2. USAID dismantling: Shut down the agency that administered most U.S. foreign aid, firing 83% of staff. This destroyed implementation capacity built over decades. A Lancet study estimates 14 million excess deaths by 2030 from the cuts.
3. Schedule F: Reclassifying ~50,000 career civil servants to strip job protections, enabling political firings. This converts expertise-based positions into loyalty-based ones.
4. Firing independent agency heads: Removed FTC commissioners, NLRB members, MSPB chair. DOJ sent a letter to Congress stating it would no longer defend the constitutionality of for-cause removal protections—explicitly signaling intent to overturn Humphrey's Executor, the 1935 precedent that lets Congress create agencies insulated from presidential control.
5. Unqualified appointments:
   • Robert F. Kennedy Jr. (HHS): Longtime anti-vaccine activist with no public health administration experience. His department's official MAHA report cited nonexistent studies, then introduced additional errors when revised.
   • Pete Hegseth (Defense): Fox News weekend host with no Pentagon or senior military experience. Participated in the Signalgate breach, sharing military operation timing on an unsecured group chat that accidentally included a journalist.
   • Kash Patel (FBI): No law enforcement leadership experience. Authored a book, Government Gangsters, containing a list of government officials he considers enemies of Trump, and has stated publicly he intends to pursue them.
   • Tulsi Gabbard (DNI): Former congresswoman with no intelligence community experience. Sen. Dick Durbin said she "would not be qualified for an entry-level position" at an intelligence agency.
   • Betsy DeVos (Education, 2017): Billionaire Republican donor with no education policy experience. First Cabinet nominee in history to require the Vice President to break a tie vote for confirmation.

This is a small sampling; a comprehensive account would be far longer.

For example, what kind of coalition would be able to actually update that Trump is less bad than it thinks if your current fears don't come to pass?

The things that have already come to pass are already deeply damaging to our institutions. We are not in 2016 when we could only extrapolate how he would behave as president. We have a decade of evidence since then. He may not play out every single thing people fear, but, like, if he doesn't end up trying to acquire Greenland by force, that's not much of an update. Not everything worth worrying about comes to pass. It's absurd that such a thing is plausible for him, even if it's not probable.

My first thought was Amazon leveraging this for drone delivery.

It's more of a continuum than that. For example, some people argue that Claude Code writing most of the code that goes into improving Claude Code is an early form of partial RSI. Clearly, the loop is incomplete, but Claude is accelerating his own advancement in a real way. How far out of the loop do humans need to be for it to fully count?

How about these?

2. A dense ball of electrons with no confinement or a blob of molten tungsten not exchanging energy in a bath of liquid helium.
3. A spacecraft indefinitely "parked" at an unstable equilibrium (L1/L2 Lagrange points) without any course corrections.
4. Revive Julius Caesar exactly as he was, including his memories.
5. Land a rocket upright on a 1-meter target on Mars with zero controls/sensors/guidance. Everything depends on perfect initial calculations.

Seems plausible, but I would put this under the category of not having fully solved continual learning yet. A sufficiently capable continual learning agent should be able to do its own maintenance, short of a hardware failure.

Depending on how it's achieved, it might not be a matter of maintenance/hardware failure so much as compute capacity. Imagine if continual learning takes similar resources to standard pretraining of a large model. Then they could continually train their own set of models, but it wouldn't be feasible for everyone to get their own version that continually learns what they want it to.

I consider Claude's "taste" to be pretty good, usually, but not P90 of humans with domain experience. I'd characterize his deficiencies more along the lines of a lack of ability to do long-term "steering" at a human level. This is likely related to a lack of long-term memory and hence the ability to do continual learning.

Is this sufficient? I don't really know the best place to put a disclosure.

https://en.wikipedia.org/wiki/User_talk:Alexis0Olson/Multilayer_perceptron#LLM_Disclosure

Claude Code is excellent these days and meets my bar for "AGI". It's capable of doing serious amounts of cognitive labor and doesn't get tired (though I did repeatedly hit my limits on the $20 plan and have to wait through the 5-hour cooldown).

I spent a good chunk of this weekend seeing if I could get Claude to write a good Wikipedia article if I told it to use the site's rules and guidelines and then let it iteratively critique and revise against those guidelines until the article fully met the standards. I wrote zero of the text myself, though I did paste some Q&A back and forth to NotebookLM to help with citations and had ChatGPT generate an additional flowchart visual to include.

After getting some second opinions from Gemini and ChatGPT, I will have Claude do a final round of revisions and then actually try to get it on Wikipedia. I will share the link here if it gets accepted--I don't really know how that works, but I bet Claude can help me figure it out.