The first thing Musk removed was grades. Then subjects. Then the age-based progression that has organized Western schooling since the Prussian reforms of the 1800s.

This was not iconoclasm for its own show. It was first-principles thinking — the method borrowed from physics that asks: what is actually true here, as opposed to what has simply been inherited? Applied to a classroom, the question becomes brutal: What does a child actually need? And what are we doing only because it's always been done?

The answers at Ad Astra were STEM-heavy but not STEM-simple. Students didn't study math in isolation. They encountered mathematics as a tool inside engineering problems. Ethical dilemmas were curriculum. So were logistics, economics, and human behavior modeling — all at once, inside a single exercise. Students once calculated optimal locations for Tesla Supercharger stations. Geography, strategy, economics, and behavioral modeling folded into one problem no textbook had ever posed.

The signature exercise was called Conundrums — layered problems designed to resist resolution. Not word problems. Scenarios requiring students to weigh competing values, synthesize across domains, and sit with difficulty rather than eliminate it. The point was not the answer. The point was learning to think when no answer is clean.

Students also designed and traded using their own currency, Astra — a student-made economy running inside a school building inside a company aiming for Mars. The layers of ambition were not accidental.

The school had no website. No admissions office in any recognizable form. It existed in deliberate obscurity — a stealth-mode startup applied to childhood.

Before Ad Astra, Musk's children attended Mirman School in Los Angeles — a private institution with a reported IQ threshold of 145 for admission. Mirman is rigorous and respected. Musk found it insufficient.

His critique was not original. It had been made before — by Maria Montessori in 1907, by John Dewey in 1916, by A.S. Neill at Summerhill in 1921, by the founders of Sudbury Valley School in 1968. Each argued, in different registers, that conventional schooling suppresses the capacities it claims to develop. That compliance is not intelligence. That memorization is not understanding. That the factory model of education produces people suited for a factory world — and that world is gone.

What Musk added was not a new argument. He added capital, a rocket company, and Josh Dahn — a teacher recruited from his children's previous school — to run an experiment most reformers could only theorize about.

Ad Astra shares DNA with these traditions. The emphasis on self-direction echoes Montessori. The absence of grades recalls Sudbury Valley, where students govern their own learning entirely. The project-based structure runs in the line of Dewey's progressive education — learning by doing, not by memorizing.

But Ad Astra is not Montessori. It is not Sudbury. It is more structured than unschooling, more technology-weighted than any progressive tradition, and more explicitly pointed toward a specific future — one shaped by AI, space, and engineering.

That specificity is the problem and the point.

Josh Dahn is the figure most accounts skip past. He is not a celebrity. He does not announce rockets. He designed the curriculum.

Recruited from the school Musk's children previously attended, Dahn ran Ad Astra from its founding. He shaped the Conundrums methodology. He made the pedagogical choices that turned a billionaire's dissatisfaction into an actual school. When Ad Astra became Astra Nova during the COVID-19 pandemic — shifting online, opening beyond the SpaceX campus — Dahn's fingerprints were on the transition.

This matters because it complicates the story. Ad Astra is often narrated as Musk's school, Musk's vision, Musk's bet. That framing is convenient and mostly wrong. The educational thinking came from an educator. What Musk provided was the resource and the permission to ignore convention entirely.

In 2020, the pandemic forced schools worldwide into online formats. Most institutions struggled. For Ad Astra, the crisis became a catalyst. The school moved online as Astra Nova, preserving Conundrums and collaborative problem-solving, now reaching students outside the SpaceX orbit.

The transition tested the core hypothesis: was the model portable, or was it inseparable from its original conditions — the proximity to engineers, the culture of a working rocket campus, the informal learning that happens when children walk past actual spacecraft?

The answer was partial. Astra Nova worked. The model translated. But something was lost — the intangible friction of place, the serendipitous learning that no platform replicates.

Synthesis is what happens when you try to give the Ad Astra philosophy to everyone.

It grew from a course developed at the original school. It became a standalone platform. Students engage in simulations — collaborative, game-based, high-stakes problems requiring teams to make decisions under uncertainty. The design borrows from game development. The goal is what the platform calls future-ready skills: resilience, strategic thinking, collaboration under pressure.

Synthesis does not try to replace school. It positions itself as a supplement — a place where the gap between subjects gets bridged through applied problem-solving. Available to anyone with an internet connection and a subscription.

The scalability is real. The original Ad Astra served fourteen students. Synthesis serves thousands. The model has escaped the rocket factory.

But the criticism is also real. Tech-heavy, elite-focused learning does not serve all students equally. Not every child thrives in high-intensity problem-solving environments. Many need what Synthesis implicitly devalues: arts, humanities, unstructured social time, physical movement, the slow accumulation of disciplinary knowledge that cannot be gamified.

There is also a structural contradiction. Synthesis assumes students arrive with foundational knowledge — the calculus behind the bridge, the philosophy behind the ethical dilemma. That foundation has to come from somewhere. If it comes from a conventional school system that Synthesis quietly critiques, then the platform depends on the thing it claims to surpass.

You cannot reason ethically about artificial intelligence without philosophy. You cannot engineer without mathematics. You cannot understand where you are going without history. Synthesis presupposes all of this. It does not provide it.

In 2024, Ad Astra announced a new physical campus in Bastrop, Texas — a small city southeast of Austin that has become a center of gravity for Musk's enterprises. SpaceX operations. The Boring Company. Now a school.

The Bastrop campus initially targets children aged three to nine: thirty lower elementary students, eighteen primary. First-year tuition subsidized to match local private school rates — a gesture toward accessibility that does not exit the structures of private education.

The school will be licensed through Texas Health and Human Services. It will emphasize self-directed, inquiry-based learning — not explicitly Montessori but sharing its architecture. The long-term aim is K-12.

Behind the campus: a foundation reportedly seeded with $100 million. Behind that: plans for a university. One geographic region. One industrial campus. One school. One planned university. One vision.

This configuration has historical precedent. The company towns of the industrial era co-located work and life under single ownership. The university-industry complexes that built Stanford and shaped Silicon Valley merged research and capital under mutual agreement. The ancient academy model placed learning beside production by design.

Whether Bastrop becomes a genuine educational ecosystem or a controlled environment optimized for one man's vision is not yet determined. The choices haven't been made. The outcomes aren't visible. What is visible is the scale of the ambition and the absence of public accountability that normally governs schools.

Every culture that oriented itself toward the stars also had to solve the same problem: how do you pass the knowledge on?

The Babylonians who charted the heavens in the second millennium BCE trained scribes in cuneiform schools. Students copied tablets for years before earning original thought. Precision and tradition were the curriculum. The method encoded the worldview.

The Greeks built the academy and the lyceum. Socrates asked questions until his interlocutors discovered what they didn't know. The method was the message: questioning is thinking.

Islamic scholars of the ninth through thirteenth centuries built madrasas that held theology, philosophy, mathematics, and astronomy inside the same walls. Synthesis was the curriculum. Al-Khwarizmi worked there. So did Ibn Rushd.

The navigators of Polynesia crossed thousands of miles of open ocean using star paths and wave patterns memorized through oral tradition and embodied practice. Their curriculum was intergenerational, tactile, and unwritable. It died when the chain broke.

In each case, the method of teaching encoded a civilization's theory of intelligence. The Babylonian school said: precision is intelligence. The Socratic method said: questioning is intelligence. The madrasa said: synthesis is intelligence. Polynesian navigation said: embodied knowledge is intelligence.

Ad Astra says: problem-solving under uncertainty is intelligence.

That is a real answer. It is not a complete one. Every previous civilization that reduced intelligence to a single register eventually encountered the problems that register could not solve.

The Babylonians could chart stars but not question their kings. The Socratic tradition produced philosophers who struggled to govern. The madrasas eventually narrowed. The Polynesian tradition was erased by contact.

What Ad Astra has not yet answered — what it may be structurally incapable of answering — is what happens when the engineering tools meet a problem that requires a kind of knowing no simulation can teach.

The most persistent criticism of Ad Astra is also the simplest: exclusivity.

From 2014 through its various incarnations, the school has served children within Musk's social and professional orbit. You could not apply to Ad Astra the way you apply to a public school. Attendance was proximity to power. No pedagogy changes that fact.

This matters beyond fairness. Educational experiments conducted in controlled, well-resourced environments with self-selected, highly motivated families tell us very little about what works for everyone. The history of education is full of models that succeeded in those conditions and collapsed everywhere else.

There is also the question of who bears the downside. Musk's model removes grades, abandons conventional subjects, and builds learning around projects rather than content areas. For some educators, this is necessary disruption — traditional schooling has resisted change for over a century and produces graduates skilled at tests, not thinking. The mold needs breaking.

For others, the absence of structure is a risk borne by children, not by their designer. Students who spend formative years in an ungraded, project-based environment may develop remarkable problem-solving skills. They may also lack the academic credentials, content knowledge, and institutional familiarity needed to navigate higher education and professional life as those institutions currently exist.

Musk's children are insulated from this risk by wealth and connection. The child who adopts the model on faith is not.

This is not an argument against experimentation. It is an argument for honesty about who experiments on whom.

The alternative education traditions that preceded Ad Astra have track records. Montessori has decades of evidence, across income levels, in diverse settings. Sudbury Valley has longitudinal studies of graduates. Democratic education has a documented history of outcomes — some encouraging, some cautionary. Ad Astra's deliberate obscurity and its reluctance to locate itself within any of these traditions has made it easier to mythologize and harder to evaluate.

That obscurity may have been strategic. Or it may simply reflect the priorities of a man more interested in Mars than in publishing outcome data.

Here is what is established: Ad Astra removed grades, designed problem-based curricula, recruited real educators, and produced students who reportedly wanted to return from vacation to continue learning. The model generated Astra Nova, then Synthesis, then a planned K-12 campus in Texas.

Here is what is debated: whether the model generalizes beyond wealth and self-selection, whether STEM-weighted problem-solving develops the moral imagination complex problems demand, whether Synthesis's scalability preserves what made the original work.

Here is what is speculative: whether Bastrop becomes a genuine alternative education ecosystem, whether Musk's planned university transforms anything, whether any of this reaches the children whose education most needs transformation.

The deeper argument underneath all of it is older than SpaceX and older than Musk. It is the argument about what education is for.

If education is for knowledge acquisition — the transmission of accumulated information to the next generation — then Ad Astra is incomplete. It does not systematically teach history, literature, philosophy, or foreign languages. It bets that a student who can solve problems can acquire knowledge when needed. That bet is unproven at scale.

If education is for capacity development — building the cognitive and moral tools to navigate a world that changes faster than any curriculum can track — then Ad Astra is at minimum a serious attempt. It teaches students to sit with complexity, reason across domains, and treat failure as information. In a world shaped by artificial intelligence and climate disruption and interplanetary ambition, those capacities are not optional.

The phrase ad astra carries a longer form: per aspera ad astra. Through hardship to the stars. The hardship is not incidental. It is the method.

The hardest questions about education are not about grading systems or classroom design. They are about who children are, what they are capable of, and what world they are being prepared to build. Those questions belong to everyone. Every public school board, every underfunded classroom, every parent calculating what their child needs — they are all running the same experiment Ad Astra ran, with fewer resources and no billionaire's foundation to absorb the loss if the experiment fails.

Self-governance in education means not waiting for Musk or any other patron to solve it. The tools are available. The evidence is partial but real. The alternative traditions exist and have track records. The window for building something that works for more than fourteen children in a rocket factory is not infinite.

Build now.