Astral Codex Ten

SpaceX

47 min read

SpaceX

Article

SpaceX is a recurring organization in the Astral Codex Ten archive, appearing 20 times across 20 issues between August 25, 2021 and March 03, 2026. The archive places it in contexts such as “spacex-launch-puts-out-much-co-flying-people-across-atlantic”; “SpaceX is even better than however good it was you thought last time you read an article about it”; “Elon Musk said the odds of SpaceX working were “less than 10%”“. It most often appears alongside Elon Musk, California, Tesla.

Metadata

  • Category: Organizations
  • Mention count: 20
  • Issue count: 20
  • First seen: August 25, 2021
  • Last seen: March 03, 2026

Appears In

Source Context

Recovered passages from the original issue text. When the raw archive preserved outbound links inside the source passage, they are listed directly under the quote.

Carbon Costs Quantified
August 25, 2021 · Original source
14. https://www.treehugger.com/spacex-launch-puts-out-much-co-flying-people-across-atlantic-4857958
Links For September
September 20, 2021 · Original source
15: Speaking of progress, every few months I see an article saying that however good you thought SpaceX was before, it’s even better than that. Anyway, according to this article, SpaceX is even better than however good it was you thought last time you read an article about it. Lots of stuff there, but one key point is that Starlink is pretty close to becoming a completely uncensorable Internet service. Elon Musk is apparently aware:
Inline links: this article
Plausibly SpaceX will still have to follow the rules of its host country, the US. “The whole world will depend on US internet regulatory law” would have sounded more inspiring a few years ago than it does today. But this Tumblr user gives the sci-fi ending I was looking for:
Inline links: this Tumblr user
Book Review: The Scout Mindset
September 29, 2021 · Original source
Galef says not necessarily. Did you know that Jeff Bezos said outright he started off with a 30% chance Amazon would succeed, even going so far as to tell investors “I think there’s a 70% chance you’re going to lose all your money”? Or that Elon Musk said the odds of SpaceX working were “less than 10%”? Ethereum founder Vitalik Buterin said he’s “never had 100% confidence in cryptocurrency as a sector…I’m consistent in my uncertainty”. And since the book came out, I stumbled on this profile of billionaire Sam Bankman-Fried, which says he believed his chances of success “were only 20% to 25%”.
Inline links: this profile
21
November 01, 2021 · Original source
(source, units are billions of dollars) AFAIK, right now SpaceX is worth about $100 billion. But the median estimate for 2030 is $500 billion. An 8% rate of return over nine years is ~100%, so even in a great economy the average company will “merely” double by then, whereas SpaceX will quintuple. Seems bold to say a company is undervalued by a factor of >2. I guess this doesn’t technically violate any theorem about stock markets or prediction markets because SpaceX is a private company. Maybe $100 billion is its valuation by normal private investors, and $500 billion is what the sort of people who buy Tesla stock would give it, and Metaculus is siding with the Tesla buyers? Still, take it public!
Inline links: source, is worth about $100 billion
AFAIK, right now SpaceX is worth about $100 billion. But the median estimate for 2030 is $500 billion. An 8% rate of return over nine years is ~100%, so even in a great economy the average company will “merely” double by then, whereas SpaceX will quintuple. Seems bold to say a company is undervalued by a factor of >2. I guess this doesn’t technically violate any theorem about stock markets or prediction markets because SpaceX is a private company. Maybe $100 billion is its valuation by normal private investors, and $500 billion is what the sort of people who buy Tesla stock would give it, and Metaculus is siding with the Tesla buyers? Still, take it public!
Inline links: is worth about $100 billion
Links For June
July 01, 2022 · Original source
37: People say that US environmental impact regulations are onerous, but all SpaceX has to do in order to keep launching rockets from its Texas spaceport is all the usual stuff, plus hire a biologist to investigate the effect of lighting on sea turtles, plus perform quarterly cleanups of local beaches, plus operate an employee shuttle, plus write a report on the Mexican-American War (really!), plus “install missing ornaments” on a local historical marker, plus help protect ocelots, plus make an annual donation to a state recreational fishing program, plus ~70 other things.
Inline links: all SpaceX has to do
Billionaires, Surplus, And Replaceability
August 31, 2022 · Original source
I don’t know what a fair economic system that takes this model into account would look like. It wouldn’t look like “billionaires shouldn’t exist”, because it might be that pushing Amazon forward by two years is indeed worth several billion dollars. It wouldn’t even look like “higher taxes on all billionaires in the hopes of having them have less money and maybe that will bring them closer to the fair level”, because plausibly some billionaires aren’t replaceable - I’m not sure anyone else would have started SpaceX if Musk hadn’t. But this model convinces me that “taxing billionaires a lot” and “taxing billionaires not at all” are at least two different unfair failure modes with their own advantages and disadvantages from a desert point of view.
Highlights From The Comments On Billionaire Replaceability
September 22, 2022 · Original source
Now that SpaceX has actually *done it*, the respected experts are changing their tune, and the investors are following suit.
But Northrop Grumman was never going to "fill that natural niche", because Northrop Grumman is run by a committee which will follow expert opinion and pre-SpaceX expert opinion was that the niche didn't exist. Blue Origin has been around for twenty-two years and still hasn't managed even an expendable orbital launch system. And the rest, could never have afforded it.
If he'd been capped at or taxed into oblivion at say half a billion, lest we ever have any unsightly billionaires in our economy, SpaceX would just be another small launch provider.
Even More Bay Area House Party
January 04, 2023 · Original source
“It’s not just about shorting Tesla. This is the beginning of a whole new antifinance revolution. Lots of people want to invest in SpaceX, but they can’t, because Elon Musk selfishly refuses to go public. No one else can get around that. But we can! We print a million shares of synthetic SpaceX stock and a million shares of antistock, each antistock share requires you to pay one one-millionth of SpaceX’s yearly profits to the holder of one stock share. Or, you know how millions of ordinary people can’t afford homeownership anymore because Blackrock keeps buying up all the homes as investment properties? We just print a million synthetic houses and a million antihouses, where the antihouse owners have to pay the average rent in a certain area to the synthetic house owners each month. Blackrock gets to invest in real estate without having to worry about all those boring contingent things like mold or termites, and we can leave the real physical houses for ordinary families.”
Grading My 2018 Predictions For 2023
February 20, 2023 · Original source
1. Widely accepted paper claims a polygenic score predicting over 25% of human intelligence: 70% 2. …50% or more: 20% 3. At least one person is known to have had a “designer baby” genetically edited for something other than preventing specific high-risk disease: 10% 4. At least a thousand people have had such babies, and it’s well known where people can go to do it: 5% 5. At least one cloned human baby, survives beyond one day after birth: 10% 6. Average person can check their polygenic IQ score for reasonable fee (doesn’t have to be very good) in 2023: 80% 7. At least one directly glutamatergic antidepressant approved by FDA: 20% 8. At least one directly neurotrophic antidepressant approved by FDA: 20% 9. At least one genuinely novel antipsychotic approved by FDA: 30% 10. MDMA approved for therapeutic use by FDA: 50% 11. Psilocybin approved for general therapeutic use in at least one country: 30% 12. Gary Taubes’ insulin resistance theory of nutrition has significantly more scholarly acceptance than today: 10% 13. Paleo diet is generally considered and recommended by doctors as best weight-loss diet for average person: 30% 14. SpaceX has launched BFR to orbit: 50% 15. SpaceX has launched a man around the moon: 50% 16. SLS sends an Orion around the moon: 30% 17. Someone has landed a man on the moon: 1% 18. SpaceX has landed (not crashed) an object on Mars: 5% 19. At least one frequently-inhabited private space station in orbit: 30%
Book Review: Elon Musk
September 13, 2023 · Original source
Musk creates cognitive dissonance: how can someone be so smart and so dumb at the same time? To reduce the dissonance, people have spawned a whole industry of Musk-bashing, trying to explain away each of his accomplishments: Peter Thiel gets all the credit for PayPal, Martin Eberhard gets all the credit for Tesla, NASA cash keeps SpaceX afloat, something something blood emeralds. Others try to come up with reasons he’s wholly smart - a 4D chessmaster whose apparent drunken stumbles lead inexorably to victory.
Elon Musk: Tesla, SpaceX, And The Quest For A Fantastic Future delights in its refusal to resolve the dissonance. Musk has always been exactly the same person he is now, and exactly what he looks like. He is without deception, without subtlety, without unexpected depths.
Inline links: Elon Musk: Tesla, SpaceX, And The Quest For A Fantastic Future
His employees seem to think so. Here’s a quote from former SpaceX employee Kevin Watson:
Highlights From The Comments On Elon Musk
September 18, 2023 · Original source
And on the workload front I used to put in 70+ hour weeks every once in a while, so working for SpaceX would be a large step up in quality of getting things done and getting to build the cool stuff, while not being a horrendous downgrade in the other dimensions (though 75 hour weeks eat you alive. It’s basically 6 hours of sleep per night and every other moment is reserved to working or getting ready for work / commuting).
I have worked in space engineering for years and my impression is that the big space agencies and companies have a lot of inertia and reluctance to consider new ideas and change. A lot of the processes have been built up in response to past failures, but they also stifle a lot of innovation. When people come in with a fresh approach and the resources to implement them, they have tended to get quite far. You can look at how SpaceX has done, but also the early days of the space program at NASA were a lot more open to innovation than today.
I remember the days when SpaceX was really ramping up university recruitment. They were the table at the career fair everyone wanted to give their resume to. Naturally, SpaceX sent a spectacular a-hole who yelled at and belittled most of the students applying. It got so bad they actually apologized about it when they held a talk at the next career fair. Turned quite a few folks off, it was a real embarrassment. Took a couple of years to wash that one out.
Open Thread 295
September 25, 2023 · Original source
3: Comments of the week: does SpaceX outperform typical aerospace contractors because of the difference between agile and waterfall engineering? And are Musk’s mental health issues related to chronic pain?
Inline links: does SpaceX outperform typical aerospace contractors because of the difference between agile and waterfall engineering, are Musk’s mental health issues related to chronic pain
What Ever Happened To Neoreaction?
December 07, 2023 · Original source
Progress Studies: Part of the appeal of neoreaction was that the past seemed better at a lot of practical and important things than the present. The 1950s gave us moon missions, the interstate highway system, cheap housing, amazing public infrastructure, and ambitious government programs to end poverty. Nowadays NASA struggles to launch anything without help from SpaceX, the government is too gridlocked for Congress to pass even small tweaks, and the tiniest amount of new infrastructure costs billions and suffers decades-long delays.
In Continued Defense Of Non-Frequentist Probabilities
March 21, 2024 · Original source
What is the probability that AI will destroy humanity this century? The argument against: usually we use probability to represent an outcome from some well-behaved distribution. For example, if there are 400 white balls and 600 black balls in an urn, the probability of pulling out a white ball is 40%. If you pulled out 100 balls, close to 40 of them would be white. You can literally pull out the balls and do the experiment. In contrast, saying “there’s a 45% probability people will land on Mars before 2050” seems to come out of nowhere. How do you know? If you were to say “the probability humans will land on Mars is exactly 45.11782%”, you would sound like a loon. But how is saying that it’s 45% any better? With balls in an urn, the probability might very well be 45.11782%, and you can prove it. But with humanity landing on Mars, aren’t you just making this number up? Since people on social media have been talking about this again, let’s go over it one more depressing, fruitless time. 1. Probabilities Are Linguistically Convenient I think everyone agrees it’s meaningful and useful to say things like “Humanity probably won’t land on Mars before 2050”. That is, suppose NASA and ESA and SpaceX hire a team of experts to calculate whether they can make it to Mars by 2050. They examine all the evidence and find that going to Mars is much harder than anyone thinks, all the rockets that people plan to use for the task are fatally flawed, and it would take decades to invent better ones. They don’t come up with a mathematical model or form a distribution, they just intuitively notice all these things, and how they add up. When giving her report, the leader of the team says “We don’t think humanity can land on Mars before 2050.” This person hasn’t done anything wrong - it’s impossible to communicate useful information without doing something like this. How unlikely is it? Again, it seems like there might be different degrees of unlikelihood. For example, it might be that it’s pretty hard to make it to Mars by 2050, but with a strong effort and very good luck, we could manage it. Or it might be that it’s insane to even consider that we could make it to Mars by 2050, there are twenty unsolvable problems in the way and everyone who claims to be working on them is a total fraud. So probably the leader of the team should be allowed to say either “It’s pretty unlikely we can make it to Mars by 2050” or “It’s very unlikely we can make it to Mars by 2050.” Suppose she says “it’s very unlikely”. We might still want to know more information. It’s “very unlikely” humans can make it to Pluto by 2050, and also “very unlikely” we can make it to the Andromeda Galaxy by 2050, but these seem like different levels of unlikeliness. You can sort of imagine a scenario where everything goes right and we make it to Pluto, but the Andromeda Galaxy would require totally new science that suspends apparently ironclad physical law. So maybe we need at least two levels of “very unlikely” - one corresponding to the likelihood of getting to Pluto, and one corresponding to the likelihood of getting to Andromeda. We could call these “very unlikely” and “extraordinarily unlikely bordering on impossible”. How many terms like “slightly unlikely”, “very unlikely”, “extraordinarily unlikely”, etc do we need, and how will we make sure that everyone knows what they mean? It seems like having these terms is strictly worse than using a simple percent scale, where the leader of the Mars investigation team says “I think there’s about a 5% chance we can make it to Mars by 2050”. This is obviously clearer than “I think it’s unlikely” and prevents you from having to answer a bunch of followup questions. There are lots of people and space agencies who want to do different things if the chance of making it to Mars is 5% vs. 25%, and collapsing those both under “unlikely” or making people strain to figure out the meaning of “unlikely” vs. “very unlikely” and which one corresponds to 5% vs. 25% feels stupid, like deliberately introducing noise into your communication and asking people to solve a Twenty Questions game before they figure out your true opinion. To put it differently, saying “likely” vs. “unlikely” gives you two options. Saying “very likely”, “somewhat likely”, “somewhat unlikely”, and “very likely” gives you four options. Giving an integer percent probability gives you 100 options. Sometimes having 100 options helps you speak more clearly. A counterargument against doing this might be that it falsely introduces more precision than you can back up. But this isn’t true. Studies find that people who use probabilities often are well-calibrated - ie when they say something is 20% likely, it happens 20% of the time. Other studies find that when superforecasters give a very precise probability (like 23%) the extra digit adds information (ie the thing really does happen closer to 23% of the time than to 20% of the time). In this case, demanding that people stop using probability and go back to saying things like “I think it’s moderately likely” is crippling their ability to communicate clearly for no reason. 2. Probabilities Don’t Describe Your Level Of Information, And Don’t Have To Lots of people seem to think that naming a probability conveys something about how much information you have (eg “How can you say that about such a fuzzy and poorly-understood domain?”). Some people even demand that probabilities come with “meta-probabilities”, an abstruse philosophical concept that isn’t even well-defined outside of certain toy situations. I think it’s easy to prove that none of this is necessary. Consider the following: What’s the probability that a fair coin comes up heads?
Inline links: Other studies find
Bureaucracy Isn't Measured In Bureaucrats
January 09, 2025 · Original source
That’s worse! A few years ago, I debated Kevin Drum about (what I considered) a particularly egregious case where the FDA dragged its feet approving a life-saving medication. Drum argued that the FDA had behaved well. In support, he found some quotes from the doctor working on the medication, who praised all the FDA bureaucrats she had interacted with, calling them extremely helpful. This bothered me for a while, until I realized that of course it was true. In the model above, each bureaucrat processes ten forms. If the bureaucrats are benevolent, this might look like talking to the doctors, walking them through the process of figuring out their ten forms, and doing the work to add their ten forms to the FDA’s growing pile of evidence supporting the application. All of this co-exists comfortably with the insight that making doctors fill out a thousand forms before they can use a medication is an impediment to medical progress. This really sunk in for me when I read an article about the fall of Afghanistan to the Taliban in 2021. Many Afghans had collaborated with the Americans, eg as translators, in exchange for a promise of US citizenship. As the Taliban advanced, they called in the promise, begging to be allowed to flee to America before they got punished as traitors. The article focused on a heroic effort by certain immigration bureaucrats, who worked around the clock with minimal sleep for the last few weeks before Kabul fell, trying to get the citizenship forms filled in and approved for as many translators as possible. It made an impression on me because nobody was opposed to the translators getting citizenship, and the bureaucrats were themselves the people in charge of approving citizenship applications, so what exactly was forcing them to go to such desperate lengths? If you ponder this question long enough, you become enlightened about the nature of the administrative state. If you don’t, you end up like Ramaswamy, who seems to think that halving the number of bureaucrats will halve the number of forms that need to be filled out. I think in his worldview, the FDA will think “Now that we have fewer bureaucrats, it would take forever to complete our current process, so let’s simplify the process.” Maybe he is working off a thesis where red tape expands to consume the resources available to it (as measured in bureaucrats). But my impression is that the amount of red tape is determined more by things like: — How likely is it that their decision will get challenged in court? And if it gets challenged in court, what amount of paperwork do they have to show the judge to prove that they made the decision on a “reasonable basis”? For example, when I type “FDA sued” into Google, the top result is a news story from a few days ago, saying that an environmental organization sued the FDA for not listening to their earlier request to ban phthalates from food. Six years ago, the environmental groups submitted a petition (the catchily-named “Food Additive Petition 6B4815”) demanding that the FDA ban 28 phthalates. Two years ago, after consulting with industry, the FDA finally banned 23 phthalates but said that the other five were okay, releasing a 58 page decision explaining its decision. Two days ago, the environmental groups sued, saying the remaining 5 phthalates are still bad. I assume the lawsuit will nitpick the details of the the 58 page decision, trying to prove that it it didn’t violate any of hundreds of federal laws saying that bureaucratic decisions must be reasonable, bureaucratic decisions must be based on science, bureaucratic decisions must respond to the petitioners’ complaints, bureaucratic decisions cannot have disparate impacts on different races, etc. I also assume that if the FDA had banned all the phthalates, they would have faced an equally serious lawsuit from Big Phthalate saying they were unfairly crippling business. Why does it take six years to respond to a petition? My guess is because they knew they would get sued and so they have some sort of million-step process that addresses every single thing you can sue over, so that they can prove to the court that their process addresses all possible complaints and they followed it to the letter. If you cut their bureaucrats in half, that doesn’t mean there will be fewer steps in the process. It means they’ll keep wanting not to get sued, the process will stay the same, and everything will take twice as long. — What has Congress mandated that they do? For example, when I Google “Congressional FDA mandate”, I get a page on HR 7248, a bill currently making its way through Congress, which says: This bill requires the Food and Drug Administration (FDA) to establish a process that supports nonclinical testing methods for drug development that do not involve the use of animals. Specifically, the FDA must establish a pathway by which entities may apply to have nonclinical testing methods approved for use in a particular context. Qualifying methods must be intended to replace or reduce animal testing and to either improve the safety and efficacy of nonclinical testing or reduce the time to develop a drug. The FDA must issue its decision within 180 days of receiving an application. The FDA must also prioritize the review of applications for drugs that are developed using an approved nonclinical testing method. The FDA must annually post a report on its website that summarizes the results of the bill's implementation, including the number of applications received, types of methods that were approved, and the estimated number of animals saved as result of these methods. So the FDA has to establish this process and post an annual report on its website. How many bureaucrats per year does this take? Maybe five? If you halve the number of people at the FDA, you still need a constant five bureaucrats to comply with this particular law. If the bill passes, the FDA comes up with a nonclinical testing process, and someone (eg the nonclinical testing industry) doesn’t think it’s good enough, they can sue the FDA for not following the law. How good a nonclinical testing process will the FDA need in order to avoid lawsuits under this bill? I assume there is a large body of administrative law answering that question, and that it will take many bureaucrats to figure this out. Finally, I admit I’m a bit confused by this. IIRC “nonclinical testing” refers to things like testing drugs on stem cells or artificial organs instead of humans. You can obviously do this for some parts of the drug testing process, but not others; the FDA has already adjusted for this and integrated it into their guidelines to some extent. I can’t tell whether this law is a righteous attempt to correct bureaucratic foot-dragging, or a powergrab by Big Nonclinical Testing demanding that the FDA privilege their products over other forms of experiment. If the latter, the FDA may try to come up with some fake pathway that satisfies the letter of the law without really giving Big Nonclinical Testing any unfair privileges, and Big Nonclinical Testing will probably sue and say it violates this bill. How many bureaucrats do you think it will take to manage that? — How much will they get yelled at if they take too long to approve drugs, vs. if they mistakenly approve a bad drug? This is the basic determinant of all FDA drug approvals. Halving the number of FDA bureaucrats wouldn’t have literally zero effect on this balance. It would mean that approving new drugs would be delayed twice as long. This would be a little more outrageous than the current delay, and might shift an outrage-minimizing FDA director slightly in the direction of cutting rules. But solve for the equilibrium: there would still be more delay than there is now. Also, I don’t think public outrage about long drug delays is linear with regard to delay, and public outrage at bad drugs is constant and large. So I think at best, firing bureaucrats would shift this balance a small amount, and only by making everything overall worse. II. One possible objection: this assumes that the average bureaucracy is like the FDA drug approval process. But the FDA drug approval process’ job is to approve things. Maybe the average bureaucracy’s job is to ban things. Then decreasing their capacity would be good. (Vivek gets to be main example here because he tweeted, but the same considerations apply to Elon: even though the government as a whole is delaying SpaceX rocket launches, individual bureaucrats might be speeding them up through the same 1000-forms logic as in the FDA case) There’s certainly a spectrum from the most approval-focused bureaucracies to the most ban-focused bureaucracies. Thinking hard about this spectrum would be a step up from “instantly” firing 50% of all bureaucrats based on social security number. So maybe a steelman of Vivek’s point would be to fire 50% of people in the ban-focused bureaucracies (and maybe double the number of people in the approval-focused ones?) I’m still skeptical that this is how it works. The past few years have seen the cryptocurrency industry demand regulation, and the government mostly fail to step up (though crypto businesses hope the Trump administration will do better). Why do crypto businesses want to be regulated more? Because the alternative is something where it’s not clear what’s legal and anyone could be sued or shut down at any time. The chief legal officer of Coinbase, from the second link: All of us are begging for sensible standards that would allow us to get back to building great products and services and spend less time and frankly, less money, arguing over legal definitions and statutes. This isn’t because anyone specifically banned crypto. It’s because there are bans on other things (like unlicensed securities, money laundering, etc) that crypto is vaguely related to, sometimes an agency regulating these things will tell a crypto company “sorry, we think you’re illegal”, and crypto wants some specific list of things it can follow that explicitly establish it as on the right side of money-laundering and security-licensing laws. Obviously industries would prefer that these be simple and easy standards (“oh, don’t worry, you don’t have to worry about money laundering if you’re a crypto company”), but they would settle for strict regulations as long as the regulations carve out some ability for them exist at all. I’ve seen the same thing play out in another area I follow, cultured meat. There are many laws about what meat you can and cannot sell, how the animals have to be treated, what the sanitation standards are, et cetera. Some of these standards make no sense when applied to cultured meat; others, cultured meat naturally fails by default (you can’t prove you’re treating the animals in a certain way because there are no animals). Others are novel philosophical questions (can you sell cultured meat without saying it’s cultured? How big does the print need to be before it counts as saying that it’s cultured? What about on restaurant menus?) Situations like these mean that there’s no clear distinction between default-yes and default-no bureaucracies. There’s no explicit ban on crypto or cultured meat. But if you cripple bureaucracies’ ability to interact with these fields, it doesn’t mean they’re fully legal, free, and happy forever. It means they’re stuck in regulatory limbo. III. So it seems like you don’t want to fire bureaucrats, you want to cut red tape. In our toy model, you want to reduce the number of forms from 1,000 to (let’s say) 100. Then the same number of bureaucrats can get drugs approved ten times faster. In our non-toy actual model of what’s going on, this would require changing incentives. Maybe you could change judicial procedures so that fewer people sue, or the FDA needs less evidence to win any given lawsuit. This sounds hard (Vivek and Elon seem more qualified to wield chainsaws than to understand legal minutiae), possibly illegal (does the administrative branch even control how judicial procedure works?), and politically unpopular (this basically looks like telling people “f@#k you, companies can put as many phthalates as they want in food, we don’t have to prove that this decision is evidence based, and you’re not allowed to challenge us.”) Or it would require Congress to repeal legislation mandating things. These Congressional mandates are probably things that Congressmen and their constituents (either real constituents or special interests) care a lot about, so good luck getting them repealed. Also, doesn’t Congress pass like one bill per year now? This would normally make me pessimistic, but Vivek and other anti-bureaucracy activists have pointed to a recent success story: Idaho. Idaho cut their regulatory code by 38% in 2019, and since then it’s only gone down. How did they decrease red tape so fast? They did it through the power of nominative determinism. In that year, they elected a governor named Brad Little. His administration is called the Little Administration. Obviously government had to get smaller. But on a purely exoteric level, what methods did they use to pull this off? This CPAC article gives the basic story: The Little administration instituted sunset provisions that review each regulation every five years and make sure it’s justifiable.
Inline links: I debated Kevin Drum, a news story from a few days ago, a 58 page decision, have seen the cryptocurrency industry demand regulation, hope, https://substackcdn.com/image/fetch/$s_!TTSy!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fb73cf3f3-e501-49b6-b2a3-b84c602d5038_481x344.webp, This CPAC article
Your Review: Alpha School
June 27, 2025 · Original source
Same state curriculum, same worksheets, same pace. The school philosophy was “no acceleration—just go deep.” We knew this was the philosophy going in. The pitch was that instead of accelerating through the state curriculum the teachers would take their time with the kids and allow them to fully explore and master the content of each grade. When we asked for examples of what that meant in practice we were told things like: “Instead of reading more advanced vocabulary, the students will learn to read out loud and use emotion and character impressions. They will learn how to vary the timing of their reading like where and when to pause to create emotion in the listener”. That sounded reasonable! It sounded like more learning, but just different learning than what the state had mandated. In practice that was not what happened. In practice “deep” just meant “un‑measured.” Smart kids + small classes ≠ accountability. The kids had time to do music, lego building, theatre and Friday ski trips because they were all really bright. They didn’t need 6+ hours a day to learn the limited math required by the state, and since the school did not feel the need to advance faster than the state, there was no pressure to push learning at all – on anything really. There was no overall school curriculum. Every teacher did their own thing. While one first grade class had weekly spelling bees, the teacher in the other classroom did not believe in learning spelling at all. But it didn’t matter. The metrics they measured the kids on in both classes advanced “enough” that no one was concerned. Most time wasn’t spent on math or language anyway. Beyond the brochure activities like skiing and theater and the four hours of foreign language per week they split between Spanish and Mandarin (which was really a great opportunity for the kids who already spoke Spanish and Mandarin to have their egos flattered. I did not see any learning in either language class. I don’t see how you can teach a language a couple of hours a week to a group of 18 kids with skill levels from zero to fluency and expect to have any impact), a lot of time was spent on DEI. DEI was pitched as helping kids handle the emotions that often come from being sensitive gifted children (they called it “Synapse”). In practice my oldest daughter got four years of learning about the basic ideas of Martin Luther King Junior and Rosa Parks, a rough understanding that some people are non-binary, and a great deal of anxiety every time I left the water running while I was brushing my teeth. The talent drain In Spring 2024 the “intermediate-school” head resigned, as did the 40+ year veteran science teacher we had been looking forward to our daughter having, the beloved tech teacher who had built a her own proprietary “learn to type” software, plus half the lower‑school faculty. Our oldest was going to be entering fourth grade; her incoming roster read like a rebuilding year for a professional sports team. It was possible we could get her into a middle school that would feed into a top tier high school, but those did not start until 5th grade. Our best option looked like “suck it up and accept whatever we had for at least a year”. One option was to do something radical. We considered taking a GAP year and traveling the world with an organization called “Boundless” but decided the timing wasn’t right. Earlier in the year we had started exploring moving to the charter city of Prospera. There is a Montessori school there that seemed like it might be alright. And we could surround the kids with an interesting group of people (and live on the beach!). But by the spring we had ruled it out. There did not seem to be many families as part of the community and we were not comfortable with the risk profile based on what was happening with the conflict between Honduras and their charter cities. Then I stumbled across Alpha: Two‑hour mornings, life‑skills afternoons, claims of 2x learning. Marketing copy is cheap; still, the promise was different enough to warrant due diligence. The initial plan was to fly some of the kids to Austin for an Alpha summer camp for a week in June – just to try it out. But once we started exploring more my wife asked me: “Could we actually move to Austin and try it for a year? Based on what is happening at the kids' school, this might be the year to try it.” So over eight weeks we flew to Austin five times – conversations with admissions and school heads, real estate searches, kids doing shadow‑days. Every parent we spoke to was very impressed with the school. Their kids really were advancing at 2x+ speed – and no one believed it was just a “selection effect”. And every guide I spoke to was extremely impressive themselves. They reminded me of the staff you run into when visiting Disney World. They all seemed “full faced” and fully-engaged. When I asked the head of admissions how they found such good staff he told me their compensation was fully transparent. “Associate Guides” were paid $60,000/year (vs the $40,000 average for Austin teachers), “Full Guides” made $100,000 and the five “Head Guides” in the school each made $150,000. They were able to both poach the best teachers from other schools, but also bring exceptional people into teaching that would not have considered it otherwise. It also let them have very high expectations for teachers once they were hired. We pulled the trigger in July. New house. Admissions letter signed. Moving truck (plus car-mover) scheduled for October. Worst case, it would be a one‑year sabbatical from stagnation. The hypothesis I carried south Elite private school attendance buys you smaller classes, brighter kids, and fancier field trips – not academic acceleration. If Alpha was real, we’d see that differential, measurable impact by Christmas – that was when we would need to decide if we would cut bait and re-apply to schools back home (and sign the kids up for more IQ-tests. The school would not accept old ones). That prior—show me velocity, not polish—is the lens through which the rest of this review should be read. Part Two: A History of Alpha Note: This is my best attempt at piecing together the history of the school based on conversations with co‑founder MacKenzie Price, high school head Chris Locke, Alpha staff, and Alpha parents; All dates are estimates and I am SURE I have gotten some details wrong. I will come back after the fact in the comments and make corrections as I hear from the people involved with corrections. 2013 – 2017 | Garage‑School to “Alpha” MacKenzie Price, then a mortgage broker in Austin, wasn’t impressed by the city’s gifted programs. She invited a small number of neighbourhood kids (including her two) into a makeshift microschool that ran two intense, teacher‑led academic “sprints” each morning, then “life‑skills” projects after lunch. Joe Liemandt — Founder of Trinity Technology, ESW Capital billionaire and family friend (MacKenzie’s husband worked for him) — kept his own children in conventional private school until he saw the qualitative improvement in the life skills of MacKenzie’s kids. He decided he wanted his kids to join MacKenzie’s but he wanted to take the project to the next level. Sometime around 2014-2017 he joined MacKenzie as a co-founder and started writing checks. Alpha recruited more students and guides and the operation jumped from location-to-location looking for a larger permanent home. 2017 – 2020 | K-8 Expansion and 2-hour focus Alpha grew to roughly 90 students from K‑8 and stabilized. Morning “core blocks” were still teacher‑driven (20‑minute bursts, 5‑minute breaks, rinse, repeat), but focused on students engaged in exercises with rapid feedback (not lectures). Afternoon workshops covered “life skills” like how to give and receive feedback or public speaking. I have not seen academic data from this time period, but when I spoke to Chris Locke, head of Alpha’s high school (which launched around 2020), he told me the kids coming into his 9th grade program were “fine,” academically – it was their life skills, confidence, and ability to engage with adults and their peers were exceptional. At this stage no AI, no dashboard, no 2x learning, no portal — just better ratios and focused pacing and the result was well balanced kids who were enjoying their education experience (even if they were unexceptional academically). 2020 – 2022 | Platform Era Begins Somewhere along the way Liemandt hired a small engineering team to stitch together edtech learning tools. Many schools use tools like iXL, Beast Academy and Amira. Those tools fit in well with the 2-hour structured approach Alpha was using. The “platform” Liemandt’s team built was meant as a tool to free up guide time so that students could be more self-directed. The dev team stitched together the preferred off‑the‑shelf apps behind a single login, and built out tracking and dashboards so guides (and students) could easily see how they were progressing. This also gave the curriculum team (there was a curriculum team now) data to understand where students were spending their time, what tools were working, and which weren’t as effective. The Alpha Portal was born. Not only did it increase efficiency, it provided data to iterate with. Chris Locke saw the curve change incrementally: each new cohort of ninth‑graders under the new tech-enabled learning platform came in a little stronger academically. The “life skills” were now being matched by the “academic skills”. 2022 | Expansion and Iteration By having access to Alpha kids post-graduation in the high school, Locke could send feedback back to the elementary school.The kids coming out of the new program were now killing it academically on Math, Language, and Science, but they were still weak on things like History and Geography. He fed that type of information back to the curriculum designers, who iterated and improved the program. Soon, in addition to the core platform that directed students to third-party tools, the tech team was building proprietary “Alpha” tools themselves. The flagship of the in-house tools was “AlphaReads”. AlphaReads requires students to read progressively more complicated passages, followed by answering reading comprehension questions. In addition to helping the kids improve reading skills, Alpha uses it to push types of content. Instead of classes in history, geography, economics and political science, some of the reading passages will cover that material (in addition to learning how to read and understand Shakespeare and Proust). The success of the 2-hour learning platform was giving the Alpha founders confidence. Liemandt in particular wanted to see if the program had legs beyond the elite group of students being educated in Austin. Alpha’s first external test in August 2022 in Brownsville, TX – a small community on the Mexico border with less than half the per capita income of Austin. SpaceX had recently launched Starbase in Brownsville in 2014 and the employees there were not happy with the existing school options. Someone at SpaceX approached Alpha and asked if they could launch a new campus for their employees. It is unclear if any money changed hands, but when Alpha launched their Brownsville campus (available to SpaceX employees and any other locals who are interested) tuition was only $10,000 (vs $40,000 at the main Austin campus); incoming students trailed national academic standards by over a year. But after nine months on the Alpha program the first cohort of students had caught up and surpassed the national average, and they kept accelerating, achieving an average learning velocity of ~2× the national average (see section four for what that means). Brownsville was Alpha’s attempt to show that their model wasn’t just rich‑kid selection effects. Spring 2024 | Field Pilots & Ukraine Trip Alpha tuition is high for the Austin area ($40,000 vs average private school ~$10,000-$15,000), but unlike most private schools tuition is all-inclusive. There are no extra fees for computers or field trips. There are no silent auctions or appeals for donations. This “no extra fees” allows the school to do some pretty ridiculous things. In the first half of 2024 Alpha sent a group of students to Poland to help launch a 2-hour learning pilot among Ukrainian refugees. Students did not pay to go on the trip. But students also did not have a “right” to go on the trip. They had to earn it. In addition to being on top of academic and non-academic expectations, students who wanted to participate had to learn basic Ukrainian so they could interact with the students in Poland they were meant to be helping. By not linking the opportunity to payment, the school could instead link it to behavior and achievement. This year a group of kids who learned to sail during the school year are going on a sailing trip through the Caribbean – for no additional fees to the parents. I also heard that around this time Alpha began testing the 2-hour learning platform at a facility for juvenile delinquents in Florida. I heard that from one individual who was not directly involved and I have not found any written documentation on it, so unclear if it worked, it was a one off, or if it even happened. But it fits into the pattern of Alpha at this stage: “We know this program works for a specific type of kid. Let’s find out how broadly it is applicable. Can it work for everyone? Is it the solution for learning and education for the world?” Fall 2024 | “Pick‑Your‑Afternoon” Specialist Schools MacKenzie told me that there was consensus among the current parents of Alpha that the 2-hour learning program was exceptional and was making a huge difference with their kids. Their kids were all learning at breathtaking speed in a very condensed period of time. But there was NOT consensus about what the kids should be doing in the other 22-hours of the day. Some parents wanted to utilize the platform’s capabilities to go even faster. Some wanted their kids to just chill out and enjoy the rest of their day – let kids be kids. Others wanted their kids to use the freed up time to do sports, or study music. It was clear to her that “learn more faster in a short period of time” was a universal desire. But beyond that it was unclear what the “right” solution for the rest of her program was. You can make the morning ultra-personalized, but if the goal of the afternoon is socialization that you are missing in the morning, you need to have some sort of alignment on how to spend that afternoon. That challenge led to Alpha’s 2024 expansion into specialty schools. Three micro‑campuses opened August 2024: GT School (Georgetown, TX) — Alpha’s “Gifted and Talented” School. Higher admissions bar; higher academic expectations; Afternoon programming focused on excelling in “academic competitions” like chess, go, debate, public speaking, robotics, programming and Quiz Bowl.
Inline links: Boundless, Prospera, MacKenzie Price, Joe Liemandt, GT School
In Defense Of The Amyloid Hypothesis
August 14, 2025 · Original source
Anti-amyloid drugs (like Aduhelm) don't reverse the disease, and only slow progression a relatively small amount. Opponents call the amyloid hypothesis zombie science, propped up only by pharmaceutical companies hoping to sell off a few more anti-amyloid me-too drugs before it collapses. Meanwhile, mainstream scientists . . . continue to believe it without really offering any public defense. Scott was so surprised by the size of the gap between official and unofficial opinion that he asked if someone from the orthodox camp would speak out in its favor. I am David Schneider-Joseph, an engineer formerly with SpaceX and Google, now working in AI safety. Alzheimer’s isn’t my field, but I got very interested in it, spent six months studying the literature, and came away believing the amyloid hypothesis was basically completely solid. I thought I’d share that understanding with current skeptics. The ATN model The most plausible variant of the amyloid hypothesis is the A → T → N model: amyloid causes tau causes neurodegeneration. 1: Amyloid The common entrypoint, typically at least 15 years before clinically detectable symptoms [1], is accumulation of amyloid-β deposits (especially Aβ42, one of several variants). Amyloid-β is a peptide produced in healthy human beings and many other animals, probably for antimicrobial purposes [2, 3]. Factors which cause overproduction of amyloid also cause Alzheimer’s. Factors that cause decreased clearance of amyloid also cause Alzheimer’s. The clearest relationship is various genes which massively increase amyloid production (while doing nothing else); these genes are Alzheimer’s risk factors, with some of the rarer and more severe ones causing extreme versions of the disease that manifest at otherwise almost-never-seen ages. One of the clearest examples is Down syndrome, which is caused by three (rather than the usual two) copies of chromosome 21. People with Down syndrome are at much higher risk of Alzheimer’s than the general population: two-thirds will have the condition by age sixty, and 15% have it by age forty. APP, the gene for the amyloid precursor protein, is on chromosome 21. This means that people with Down syndrome will have an extra copy. This extra copy has been observed to lead to higher-than-normal amyloid levels. But there are many genes on chromosome 21; do we have additional evidence that it’s the amyloid one that’s involved? Yes. Dozens of other mutations on APP cause the same sort of extremely young and severe Alzheimer’s. So do mutations on PSEN1 and 2, the genes for the enzyme that processes amyloid precursor protein into amyloid. So do mutations on several other amyloid-related genes. [6, 91 - 96] Researchers call these autosomal-dominant Alzheimer’s, meaning Alzheimer’s cases that get inherited from a single parent in a simple fashion typical of single-gene disorders. They make up about 1% of all cases, and are our strongest evidence for the causal role of amyloid in the disorder. To my knowledge, there is no serious claim that these genes could be working through any pathway other than their shared role in the amyloid system. But these autosomal-dominant cases only make up about 1% of all Alzheimer’s patients. Might they be a different disease than the usual sporadic Alzheimer’s that strikes people without strong family histories at normal ages? Probably not: the presentation and trajectory of autosomal-dominant and sporadic Alzheimer’s cases are strikingly similar. Both show an initial appearance of amyloid pathology starting in intrinsic connectivity networks in both autosomal-dominant [14] and sporadic [15–18] types, cortical tau appearing first in the medial temporal lobe and with the exact same fold in both disease types [97] (despite human tauopathies having at least seven other possible characteristic folds [36]), that tau pathology worsening and spreading outside this region only once amyloid pathology reaches sufficient severity [65], neurodegeneration progressing closely in step with the tau pathology, and the same usual approximate trajectory of cognitive symptoms due to the sequence of affected regions. So it’s as if two bank robberies occurred hours apart, in the same town, and in a highly similar and idiosyncratic manner, and we can positively identify the culprit of one on security camera footage. It’s a good bet the culprit of the other is the same. Increased amyloid production → Alzheimer’s is an especially clear and simple pathway, but any other change in amyloid can also cause the disease. For example Overproduction or reduced clearance of amyloid due to impaired slow wave sleep. Aβ production is neuronal activity-dependent, and toxins (perhaps including Aβ) are cleared from the brain during sleep via the glymphatic system. Thus Aβ can accumulate if the brain is more active and/or has less opportunity for clearance. [7, 8, 9, 10, 11]
Inline links: Aduhelm, David Schneider-Joseph
Some years ago, I worked on the guidance and navigation software for the Falcon 9 first stage landing. The early attempts failed because of annoying issues like stuck propellant valves, running out of fin hydraulic fluid, and so on. It took many tries to iron out all those issues. Meanwhile, fans and critics kept proposing that SpaceX’s fundamental approach was flawed, and suggesting entirely different approaches. Most of that criticism was pretty uninformed.
Non-Book Review Contest 2025 Winners
October 17, 2025 · Original source
Elon Musk’s Engineering Algorithm, reviewed by a former SpaceX employee and practicing aerospace engineer who prefers to remain anonymous. He is an avid ACX reader and a published writer.
Inline links: Elon Musk’s Engineering Algorithm
25
October 28, 2025 · Original source
And the Solano Foundry would be the “the largest [advanced manufacturing] park in the US”. Many of Silicon Valley entrepreneurs’ manufacturing startups set up shop in Southern California - for example, Elon Musk’s original base for SpaceX and the Boring Company was in Hawthorne, near LA - just because the Bay has so few good industrial locations. The Foundry aims to change that, and aims for 40,000 new manufacturing jobs. Finally, something nobody else will care about but which is close to my heart - Jan is pursuing a partnership with Monumental Labs, a group working on “AI-enabled robotic stone carving factories”. The question of why modern architecture is so dull and unornamented compared to its classical counterpart is complicated, but three commonly-proposed reasons are: Ornament costs too much
Inline links: Solano Foundry, https://substackcdn.com/image/fetch/$s_!APFe!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2086d0cc-0654-4c0a-80c3-3aaacbe5744e_1199x675.jpeg, Monumental Labs, is complicated
Mantic Monday: Groundhog Day
March 03, 2026 · Original source
Third, the past few years have seen dramatic advances in financial technology. Crypto traders have invented the perpetual future, a new instrument that tracks an asset without requiring anyone to own the asset involved. That means traders can buy and sell shares of SpaceX, OpenAI, and other nonpublic companies that won’t actually give you their shares. Hedging the price of nickel used to require someone somewhere in the process to own an actual warehouse full of nickel. Now you can skip that step.
Inline links: perpetual future