Oceans Yet to Burn

This is consistent with smaller-scale issues that have been anecdotally reported by users of other models.

The unity of government which constitutes you [as] one people is also now dear to you. It is justly so; for it is a main pillar in the edifice of your real independence, the support of your tranquility at home, your peace abroad, of your safety, of your prosperity, of that very liberty which you so highly prize. ...you should cherish a cordial, habitual, and immovable attachment to [your national Union]... Citizens by birth or choice, of a common country, that country has a right to concentrate your affections. The name of American, which belongs to you, in your national capacity, must always exalt the just pride of patriotism...

There is an implicit distinction, here: the individual and the country (his society) are of different kinds.

An individual is quite limited in what he can do to secure his own liberty. If he has to fight an enemy dictator and all of that enemy's men alone, then he is likely to lose. Human beings depend on the knowledge and resources of others, including the knowledge and resources of the past, which are passed on to them; without those, a lone man can do very little.

A country can fight an enemy army successfully. Likewise, a society can create an environment within which liberty can be exercised. However, a country cannot exercise or enjoy liberty, because a country is not a person.

Washington, it would seem, believes in a relationship between citizen and country which is reciprocal and non-totalizing. The country does what a country is good at, and the citizen does what a citizen is good at, and the two support each other. Thus, a man can owe something to his society without owing everything to his society; the work of a man's hands is his own work, while society provides positioning and leverage that make that work go farther.

While video adaptations of Shirow Masamune's acclaimed 1989 manga Ghost in the Shell have tended to be more cold and philosophical (2002's Standalone Complex, for example, concerned a wave of copycat crimes without an original), judging from the first episode, this series sticks a lot closer to the manga – it's warmer, funnier, more cartoonish, and the characters are less competent – while being no less action-packed.

How is the animation quality? If you sent this episode back in time to 2003 and told me it came from the year 2026, I could believe you; it stacks up favorably against Standalone Complex, which IGN described as "the most expensive season of TV anime ever produced at the time." It's good enough that I'm tempted to look for other series by this same studio.

How is the dub? Not bad. The music? Better than 2013's ARISE. Neither quite live up to the standard set by Standalone Complex, but both the dub and the score on that show were legendary, so few things do.

How is the decision to stick closer to the manga? Time will tell (only one episode is out so far), but it's probably the right call. The presence of tube televisions is reminiscent of Netflix's 2024 reboot of Ranma ½, choosing to present a classic manga as of a time period.

For this test, please zoom in to view the image at a larger size and have a look at it before answering.

My personal feeling is that something about AI-generated images makes them uncomfortable to look at when observing the details, but I'm not sure if I'm just imagining it. I usually try to make the "AI Generated" stamp large enough so that it's always within the field of view while looking at an AI-generated image, but for this test, I have deliberately made it much smaller.

Can this problem be solved without a war?

Can a war actually solve this problem?

If this problem is solved with a war, will it stay solved?

LLM responses are (deliberately) somewhat random, but an LLM will attempt to maintain consistency with its own prior posts in a conversation. If it makes an error (as they do), then it is likely to go uncorrected. If it makes a judgment, then it will likely not depart from it. This means that the errors (and the judgments) may compound. The longer the conversation runs, the more opportunity there is for it to drift.

The LLM drawing from past conversations provides three additional opportunities for error: (1) the LLM could draw from an unrelated conversation or document (e.g. a fictional document about an SCP Foundation task force), (2) the LLM could misread the information in the conversation (as they sometimes do), or (3) the information itself (possibly at the end of a long conversation) could already be an error.

There was once a time when the United States of America was new. In 1796, when George Washington published what is now called his farewell address, 1776 was only 20 years prior, and the Constitution had only been in effect since 1789.

Things that started before we were born often feel as though they are the way the world has always been. By the start of the U.S. Civil War, a man born in 1789 would be 72, having lived a full lifetime under the Constitution; that is about how long the Soviet Union lasted.

Succession is a tricky business:

Was the system the work of one man, tied together by his unique genius, or can it support itself?

Will successors agree to maintain the process of succession, or attempt to stop it?

Can the system of succession carry its own weight without any of its original architects?

250 years is a long time. America made it past the first few successions, and then it made it past the first lifetime, and then it made it past a second lifetime, and now a third. It pre-existed the Soviet Union, and then continued on after it.

I am not entertained.

Hmn. I can't be Sam Kriss, and becoming u/KulakRevolt would be a type of failure.

Let's see what we have (>100 notes only):

Rock Island Willerbean, 2017, 5,600 notes, 92 words, political cartoon (topical issue: exotic pet trade)

Near-Future Sci-Fi Politics Chart, 2017, 248 notes, 101 words, political cartoon / futurist joke post

"Okay, but what if like instead of cars, it was golf carts.", 2018, 101 notes, 151 words, art (simple 3D rendering) / autodialogue

Duterte Speedrun, 2018, 279 notes, 950 words, futurist joke post / political satire

Go read all 5,640 posts again., 2020, 73,400 notes, 15 words, tweet dunk

Reverse Harry Potter, 2021, 229 notes, 22 words, tweet (pop fandom)

Billionaire Relocation Act, 2021, 167 notes, 62 words, political satire / futurist joke post

'art over 30 years old is right-wing', 2021, 312 notes, 45 words, political satire / provocation

'cruising the galaxy solving weekly moral puzzles', 2022, 432 notes, 16 words, tweet (pop fandom)

'guy that uses scale models for his underground windows', 215 notes, 22 words, tweet

Moon/Cyber/Cryo Prison, 2023, 237 notes, 258 words, futurist joke post / poll / philosophy problem

Re: The ‘Chillers’ Discourse, 2023, 113 notes, ~1,200 words, futurist joke post / political satire

The Swift Pill, 2024, 133 notes, 200 words, absurd joke poll

'Omelas but we made one guy the outgroup', 2025, 375 notes, 35 words, pop philosophy joke tweet

imported twitter discourse w/ pre-provided answer (that somehow gets more discourse anyway)

Diagrams can massively cut down on word count with basically no loss in content. Consider if a given paragraph could be replaced with a picture you threw together in PowerPoint.

[ from previous ]

Compared to the Release the Leopards post from 2022, for example, it's demanding too much intellectual work, not providing enough scaffolding, and not making the payoff for that work clear.

A few other notable discourse posts as examples (from Tumblr Top):

The Monster of Castle Von Wokenstein (1,300 words, 2019)

'power skeptics support privacy' (96 words, 2022)

'poverty is the most intersectional criteria' (245 words, 2023))

Thinking about what changed…

I can write more words faster since around late 2025.

While writing Now, Melt, I used LLMs to measure density and comprehensibility to manually make the text denser.

Inspired by the Analects and Dao Dejing, I had sentences or paragraphs serve more purposes simultaneously.

Increasing length and interconnectedness predates this (see 2024's 5,500-word "You Won't 'Beat Trump at His Own Game'"), but not at the same density.

Let's introduce some more rules for this experiment:

Monday Law - All posts will be queued for a specific day of the week (the return of Monday Law).

Weekly Word Limit - The weekly post limit will be ~1,200 words (about 3 of the shorter posts).

I think part of what comes across as “too many words” is a problem of “not enough grounding”. […] Your posts are all you, […] talking about yourself and your ideas instead of something outside your head, and what passes for a conclusion after a page of [mitigatedchaos theorizes airily] is shit like this: [blockquote from post] …“We can theorize more”.

I think that you would consider the following, on some level, to be "historical common sense," not requiring specific citations or examples:

[a] Elites are neither totally unified, nor totally opposed (people can gain or lose elite status; different industries/interest groups lobby and this is inherently somewhat zero-sum).

[b] Wealth is a form of power, but not the only form of power (governments can regulate, tax, and nationalize; revolutions can expropriate).

[c] Inter-ethnic conflict risk is neither 0 nor 1 (there is peace, but there is also conflict from mild (Curley Effect) to extreme (WW2, Rwanda, etc)).

[d] All else being equal, the influence of a political force scales roughly with its size (voting blocs, armies).

That's what the post is working from, and is not the problem.

The problem is that the post is too abstract. It is trying to, tactically, avoid giving specific examples in order to invite a potential opposed reader to engage it on the level of general rules, and in the act of attempting to form a general rule, notice the absence of historical common sense.

i mean, both. Jesus fuck, dude. Can you find a single person who believes that they regularly get everything you mean out of a post?

If I thought that the mechanism were irrelevant, I would not have included it. This is good editorial criticism, though.

To go over what I've gathered so far:

Posts should generally be limited to 150-350 [1] words (typical Tumblr screen length amount). [ghs] (A small handful of exceptions could stretch to 1,200 words, but no more.) [kitswulf]

Pay more attention to the surface layer of the text. If a point is important, foreground it, so that it's visible while skimming. [ghs]

Material for a close reading or second pass can be included, but don't assume a close reading. [ghs] It's essentially "bonus content."

If it doesn't fit in 350 words without a close reading, then it doesn't fit; it may be better to break it up into multiple posts, or to focus on only one branch of an argument at a time.

The post should focus mostly on one concept, rather than providing a 'wall of insights' or pursuing multiple simultaneous objectives,[2] so that it can be linked externally to quickly explain a concept. [kitswulf]

The content of what a post will cover and why should be clearly signposted, so that the reader knows not only what they can expect to get out of the post, but also what they are looking for. (Making the post shorter also makes the payoff sooner.) [bithy]

Even Substack is 'high-quality news-magazine content,' so the long essays are too long. [irradiatespace]

Some of this is similar to some experiments I've been doing offline, but I see some potential software-style dependency graph and modularity/integration issues.

[1] Ideally closer to a max of 325, so that the bottom of the post is visible on a laptop or desktop computer.