myrmidon

I just realized that heating and expelling something electrically vs just burning in a rocket engine makes absolutely no difference from an ISP perspective. All that matters is exhaust velocity, and that basically only scales with temperature and molecule size (lighter being better), so steam is no good anyway.

Hypothetically though, if you would just pressurize e.g. H2, then cool it, then pressurize it further-- should that not get you higher exhaust velocity that just heating/burning? why does that scheme not work?

What exactly prevents you from just heating the steam to like 2K? Would the ISP still be worse, or is that just more of an engineering nightmare?

To expand on this: Turbogenerators use hydrogen because it causes little drag, its cooling properties are good for a gas and it's much cheaper than helium.

Liquid cooling would be better but cooling pipes are not very amenable to spinning around at 3000 RPM and you have to keep the rotor inside the generator as compact as possible so centrifugal force won't rip things apart.

The non-rotating part of the generator is typically water cooled in bigger generators by piping the water straight through the hollow copper windings themselves (in slower rotating generator types, water cooling is also used for the spinning rotor).

You do have valid points with the "better to avoid mixed criticality when possible" and "isolate the most latency sensitive compontents on dedicated hardware".

I work in embedded software and I have some insight on why the reality often looks very different from this:

Software "cost" works differently. In a startup environment, you can spread the fixed costs for software development on thousands of installs (or even more). Automation software is very different, and solutions are often bespoke; even if you're building and selling machines, the numbers will be much lower than for consumer stuff.

This means that fixed costs for software development matter much more, and that hard developer productivity (as in: how many factories can your PLC programmers automate per year) is much more important. Vendoring dependencies is much more costly, splitting your codebase over multiple subsystems is less desirable (=> overhead for general development, update, configuration management, redundant subcomponents, slower iteration), heterogenous systems are very undesirable (because specialized employees are more expensive to manage and less flexible).

This is why much of the industry is so willing to shell out for "overpriced" PLC controllers (when freely programmable MCUs can be had for cents), proprietary frameworks and what you might perceive as obsolete garbage tooling for HMI/cloud integration: All those things help to build solutions that are "good enough" with less programmers, less training, less specialization and thus, crucially, less software-fixed-costs (that you wouldn't be able to recoup).

One other thing: software being formally correct vs. reliable/robust is much more orthogonal in practice than you might assume, and industry typically cares more about the latter. My perspective is that formalizing/certifying software is (overly) concerned with eliminating lower risk holdouts (think: proving latency bounds on memory allocation or somesuch), while "robustifying" software looks at "actual highest risk" parts first (sanity checking requirements/specifications, avoiding pitfalls for users when configuring/operating your system, fixing false positives).

If you don't hate existence, raising children can also just be seen as paying back for yours. I like that view.

You can certainly do a lot even with low end CPUs in automation (and they are certainly used a lot there).

But there's a lot of basically-desktop CPUs in industrial automation too, because those are not that expensive and very useful for higher bandwidth interfaces (think ethernet based), cycle times in the microsecond range (higher CPU clock gives you a much more generous cycle budget) and just a generally easier time doing anything where a FPU and memory helps (motion planning, driving some UI on a touch panel, etc).

See:

https://www.beckhoff.com/en-en/products/ipc/embedded-pcs/

I don't think that real-time, superhuman AI performance is ever gonna happen with <1b parameter models.

Biology had a lot of time, attempts and pressure to make this work but even the most optimized lines (something like a magpie) need more complexity (almost a billion neurons) to get a fraction of superhuman intelligence.

Results with small artificial neural nets are very limited so far, too.

Such development boards are a completeley different thing compared to RVA23-profile RISC-V CPUs.

This is like comparing an 48MHz ARM-Cortex-M0 with the CPU in your Macbook. Yes both of them are ARM cores, but the small embedded processor has very little RAM/Flash/clockspeed, no memory management unit, no cache, no modern RAM interface.

Not even the instruction set is really the same, because you (typically) won't even have floating point instructions/registers on the small MCUs, much less vector instructions.

The Milk-V Jupiter 2 mentioned in the post is basically a full-fledged Desktop CPU and it's price is not gonna be anywhere close to $2 either...

If you can capture a significant fraction of the gains made by automating lots of white-collar work, then those billions/trillions look more reasonable.

"Correct" investment volume towards early industrialization might have looked just as unreasonable without the benefit of hindsight.

I don't disagree with most of what you say, but one crucial point you miss in my view is that letting public trust in banks & "the economy" collapse is really "expensive" (in hard productivity, not just on paper).

This is what was allowed during the great depression and what we desperately tried to avoid in later crises.

Western democracies are not set up at all to let the government direct and compensate millions of workers on the turn of a dime; micromanagement by government in such a situation can be expected to be extremely wasteful and ineffective, and this only got worse, because we have less generic/unqualified use for human labor than in the past (construction, railroads, motorways, heavy industry), and specialists are even harder to manage.

If everyone stops even just believing in the ability of local supermarkets to provide toilet paper next week or, on the other end of the spectrum, starts distrusting the banking system/dollar entirely, then the consequences can be extensive, expensive and slow to recover.

It might even be worth it to feed taxpayer money to rich greedy capitalists because the public comes out on top still by avoiding that trust collapse.

Because (some) investors with money still believe that this overspending (tbd.) on R&D is gonna pay off in the form of market share (or the generosity of our future AI overlords ;), and (some) dumb money is just following it.

To be fair: You'd need to make at least several billions of revenue by selling pencil sharpeners while mostly spending the $100b on R&D (not the production of individual sharpeners)!

The market for AI is very clearly there; people are willing to pay more than enough so that providing LLM infrastructure (inference) is profitable.

Just the SOTA model R&D is pretty unsustainably expensive right now, but this only means that if the status quo "bursts" (because people realize that pushing the frontier is not paying off), model improvement will (probably) be slower afterward.

Note: this could also happen because expensive SOTA progress fails to keep ahead of "garage-trained" models, in such a case SOTA-AI lab investments collapsing would not really slow AI progress anyway.

Doomsday scenario is a self-reinforcing feedback loop of:

1) economy tanks, people earn less disposable income or become unemployed

2) consumers spending less tanks the economy even more

We have historical precedent for this in the great depression, and the government going hands-off back then made the problem almost certainly much worse.

You could even make a pretty credible argument that the Nazis only got into power thanks to this, so there's your doomsday scenario.

What I find really interesting is that black holes don't really "require" infinite density; the mass density inside the event horizon never exceeds neutron star levels, and becomes in fact lower the heavier the black hole gets (=> event horizon enclosed volume grows faster than mass).

AWS New Carlisle in Indiana is probably at ~1GW right now and scheduled to hit ~2.2GW when finished (https://www.cnbc.com/2025/10/29/amazon-opens-11-billion-ai-d...).

Microsoft Fairwater Wisconsin is currently at 400ish MW and projected to scale beyond 2GW of electrical power by mid 2028, as buildings currently still under construction come online (according to: https://epoch.ai/data/ai-data-centers/directory/microsoft-fa...).

Largest operational/planned AI datacenters are in the 2GW+ range, so ~600min for one of those.

But a more typical size would be around 200MW, which could run over a week on 20GWh.

It's around 1MWh within a fraction of a second.

That means very high peak power, making your whole capture infrastructure extremely expensive, but even capturing 100% of an average lightning bolt is worth <$100 in electricity.

Which is why nobody even attempts this.

The bare SoC you can get for ~1.20€ in Germany:

https://www.reichelt.com/de/en/shop/product/esp32-c3_soc_wif...

But even paying 5-30€ for a devkit still seems insanely cheap to me; that's like several bars of chocolate at current prices, even kids can afford that (or make someone buy it for them).

There's some plausible concepts:

https://gravity-storage.com/technical-concept/

But battery storage is a proven technology now and the cost trends with such bespoke megaprojects are not looking promising, so I'm not holding my breath.

Burning (regenerable) metal fuel in old coal power plants is also an interesting approach:

https://www.sciencedirect.com/science/article/pii/S2666352X2...

How is electricity outrageously expensive? We get power in the most flexible of forms (electrical) delivered directly to our homes at a rate that no human in history would have ever performed voluntary physical labor for (the equivalent of thousands of horses or people for a few hundred bucks an hour, assuming <500$/MWh). Even complete energy autarky has become feasible for private individuals if you are willing to invest.

Just because power used to be even cheaper in the recent past, by doing non-sustainable shit and pushing all the resulting costs onto others/the future does not mean that was ever appropriate, necessary or desirable, and everyone advocating for chasing that model is a greedy, selfish asshole in my view.

> Eating a hamburger or driving a car is not the problem at an individual level.

Strongly disagree. Oil is not getting produced because a handful of companies want to ruin the planet-- that happens because billions want those oil products, for personal mobility, cheap goods, cheap energy.

A lot of people simply don't care enough about the future consequences; they want to continue cheap consumption for as long as possible, and find justifications and excuses easily enough ("China emits so much more anyway", "I can't afford to pay more for gas", "why should I have to make any sacrifices that my parent generation did not have to make", etc.).

If you could disperse every last one of those companies into tiny 20-man operations, you would solve absolutely nothing. What would solve things is if a majority of people (in industrialized nations) would become willing to pay the cost for fast and extensive (green) electrification, to downregulate the size and frequency of personal transport, to pay a premium for green concrete and steel, to accept higher prices and worse availability in order to sanction any trading partner that is not similarly serious about emissions.

We can't get any of the above not because of evil corporations or politicians, but simply because most people just don't want to pay the price and make the sacrifice now for a better future.

This might shock you because it happens outside your social bubble, but there is an insane amount of people (double-digit percentages for sure) that is simply unwilling to make even the smallest sacrifice in order to mitigate climate change, and they vote accordingly.

That does not make climate change any less of an emergency, but it limits politically feasible countermeasures a lot.

I'm not saying that wage level is the only factor for growth. But it is extremely significant.

If you want to keep higher local wage levels, you rely on other comparative advantages (easier access to capital, better education, better political stability), and all those tend to erode over time. As they do, you either lose your wage advantage or the whole industry, there is no alternative.

> Otherwise Mexico and Russia would be industrial superpowers catching up to EU and the US, if all it took was neighbouring you and having cheaper labor than you.

They are! Just compare vehicle production for an easy proxy: Mexico had >100% growth over the last 2 decades, while US numbers are basically flat. Russia is a bad example now because of war, but they did a lot of catching up before.

If your theory is that all western manufacturing struggles just because every single company is stupid and spends too much on lawyers and accountants then I don't know what to tell you.

> You forget EU car makers have factories outside the EU, in places with much cheaper labor. As do Japan's car makers.

Yes, but many of their billions of R&D spending are still spent on (globally) very well paid (local) engineers. Not saying this is bad, but makes it harder to compete.

> Well, if we just never freed the slaves, we would have internationally competitive cotton.

The US still is the leading exporter of cotton.

> The west marches on high end innovative exports, not competing in the cheap nicknack market.

This is an overgeneralization. E.g. Several of the most successful German companies are grocery retailers (as boring and low-tech as it gets).

> Are ASML and Airbus selling well because they're cheap to make or because of their unique innovations competitors can't easily replicate?

I'm not saying you can't pay out globally uncompetitive wages as long as you're sitting on an R&D advantage, but your low-end un-innovative neighbours will catch up to you when R&D is much cheaper for them (just consider the Japanese electronics industry in the 80s/90s for example; people said the exact same shit back then until they no longer could).

That's just one perspective and I disagree with it.

European carmakers are where they are because electromobility reset their legacy advantages (a lot of which they built up more than half a century ago on much cheaper wages); now the same dollar amount gets them less R&D output than in China, because engineers are more expensive, and production itself is difficult to keep competitive for the exact same reason (also why a lot of car production has shifted east into lower wage European countries long before Chinese cars became competitive).

If wages were significantly lower because of more labor migration, the industry would (obviously) be more globally competitive as well (I'm not saying that would be desirable, just that it's a given).

Are you confident that globally uncompetitive wages are good for you long-term, though?

This might be very nice for some time, but if you "artificially" keep up local wage levels by limiting worker mobility you might set yourself up to lose the whole industry to lower wage countries (see e.g. European automotive sector).

I think you are being deceptive; the no-regulation alternative/cost to "setting up a middlemen network for responsibility diffusal" is not "bribing officials", it's nothing; companies just used to pump their toxic tailings slurry into the next lake/river/barely secured ditch at basically zero cost.

Having to pay those middlemen is a sign of the regulation working (because the cost for creating superfund sites is no longer zero), and all society has to do is increase enforcement rate/liability limits until the middlemen are more expensive than proper disposal.

Cases like this are exactly why every populist claim of "cost savings through deregulation" should be met with a healthy dose of skepticism, always.

Not only were many environmental regulations "written in blood" (and are well justified by observed damages before they existed), the average citizen would typically gain little from scrapping them. Allowing some industrialized chicken farm to dump their waste into the next river is not gonna make your eggs that much cheaper. "Socialize (external) costs, privatize gains" is often the actual play instead.

There's another approach I recently learned about: Converting coal plants to burn metal fuel (iron powder <=> rust):

See e.g.: https://www.sciencedirect.com/science/article/pii/S2666352X2...

Volumetric energy density is pretty decent, and it's easy and safe to store (albeit heavy). The big appeal is that it seems feasible to convert coal plants while touching only their burner (there is a smallscale prototype running in Germany AFAIK). You could also potentially even transport the fuel (not sure how the losses look on that though since energy/weight is not very high; maybe ships or trains could still work though).

Drawback is that regenerating the iron powder needs (green) hydrogen and the cycle efficiency is low.

But this could be very helpful for countries like Germany, Poland or especially China to get dispatchable green power from former coal power stations.

For 1.2GW in panels and 20GWh of batteries you'd currently pay about a billion for the panels and 2-3 billion for the batteries (price for a turn-key solution, not just cells). Assumptions:

- 25% capacity factor for panels

- 3 days of storage (at 300MW)

- $120/kWh for storage

- $1/W (peak) for panels

I'd argue that such a setup is a big upgrade over the reactor since you have much higher peak power; you basically get more dispatchability at a comparable capacity factor (90%-ish). I suspect maintenance to be cheaper as well.

But an actual solar setup would probably install more panels and less storage (giving you cheaper power at a lower capacity factor).

Edit: The problem in practice for the nuclear reactor is that it has to be competitive with almost the panels alone, because otherwise your consumers are just gonna go "I wont buy nuclear energy 24/7, I'll just use much cheaper solar electricity whenever the sun shines, pay maybe for a few hours of batteries and fall back to gas when I really need to (very cheap per peak MW, but not per MWh)".

If you want to be fully paid for constant power output, you have to be fully competitive at all times, not just when the sun is down.