Arion

AI • Technology • Innovation

[Arion @ar10nx]

Three failed rocket launches. A car company fighting to survive.
In 2008, Elon Musk’s biggest bets weren’t about changing the world yet. They were about surviving long enough to try.
Here’s what happened next. https://t.co/Vv7c7xiTF1

[Arion @ar10nx]

A grid battery is only useful once the work around it is finished.

At Tesla's 2025 shareholder meeting, Elon Musk describes the next installation problem: how much equipment still has to be assembled around the battery?

His answer is to move more components inside the system.

He then sets a further goal: integrate enough power electronics for direct output at up to 35 kilovolts.

That was a development target, with fewer external connections and less substation work as the payoff.

The product includes the job of getting it connected.

https://t.co/sEFraNGFY1

[Arion @ar10nx]

Three failed rocket launches. A car company fighting to survive.
In 2008, Elon Musk’s biggest bets weren’t about changing the world yet. They were about surviving long enough to try.
Here’s what happened next. https://t.co/Vv7c7xiTF1

[Arion @ar10nx]

Before asking Grok to rewrite an encyclopedia, Elon Musk says xAI first trained the model to think critically.

The model has to break an argument into axioms, test whether those premises are plausible, check for contradictions and decide whether the conclusion follows.

Only then does it move through popular Wikipedia articles and research what should be added, modified or deleted.

That sequence matters. A system that can retrieve more information is not automatically a system that can judge it well.

The ambitious part of Grokipedia is not the size of the library.

It is the attempt to make the reasoning layer visible in every entry.

https://t.co/sEFraNGFY1

[Arion @ar10nx]

Three failed rocket launches. A car company fighting to survive.
In 2008, Elon Musk’s biggest bets weren’t about changing the world yet. They were about surviving long enough to try.
Here’s what happened next. https://t.co/Vv7c7xiTF1

[Arion @ar10nx]

TANK PRESSURE HAS A COST

Starship uses its own propellants to pressurize its tanks.

In this 2024 tour, Musk explains the tradeoff: turning cryogenic liquid into hot gas adds work for the pumps.

Keeping the tanks pressurized competes with making thrust.

https://t.co/NlTDnvoCfu

[Arion @ar10nx]

Three failed rocket launches. A car company fighting to survive.
In 2008, Elon Musk’s biggest bets weren’t about changing the world yet. They were about surviving long enough to try.
Here’s what happened next. https://t.co/Vv7c7xiTF1

[Arion @ar10nx]

A rocket can lose control because something tiny freezes in the wrong place.

During the 2024 Starfactory tour, Elon Musk explains how ice clogged Starship's roll-thruster system.

The engineering problem was broader than water: material in the gas stream could turn solid at cryogenic temperatures.

The response included better strainers and valves, with redundant valve arrangements under consideration.

The aim was to keep a single valve failure from taking away the ship's ability to orient itself.

Even at rocket scale, reliability can depend on what gets through a small passage.

https://t.co/sEFraNGFY1

[Arion @ar10nx]

Three failed rocket launches. A car company fighting to survive.
In 2008, Elon Musk’s biggest bets weren’t about changing the world yet. They were about surviving long enough to try.
Here’s what happened next. https://t.co/Vv7c7xiTF1

[Arion @ar10nx]

Elon Musk’s argument is that cynicism often looks intelligent because it predicts selfishness everywhere.

Never trust a cynic to explain human motivation.

But that lens has a blind spot.

People who assume the worst in others are frequently projecting their own motives. They miss the ordinary cooperation that keeps families, companies and societies functioning every day.

Most people are neither saints nor villains. They are imperfect, mixed and—on balance—trying to do reasonably good things.

Cynicism can protect you from disappointment. It can also make you systematically wrong about people.

https://t.co/sEFraNGFY1

[Arion @ar10nx]

Three failed rocket launches. A car company fighting to survive.
In 2008, Elon Musk’s biggest bets weren’t about changing the world yet. They were about surviving long enough to try.
Here’s what happened next. https://t.co/Vv7c7xiTF1

[Arion @ar10nx]

Elon Musk used Raptor to explain the difference in 2021.

“Prototypes are easy. Production is hard.”

The engine design was extraordinarily advanced, but the problem consuming his time was not proving that one engine could work.

It was building the production system that could make enough engines, consistently, for Starship.

A prototype demonstrates possibility. Manufacturing has to survive repetition, tolerances, suppliers, inspection and pace.

That is why the factory can become the real product. An invention changes the world only when the process behind it can reproduce the result.

https://t.co/tYzuqFnY50

[Arion @ar10nx]

Three failed rocket launches. A car company fighting to survive.
In 2008, Elon Musk’s biggest bets weren’t about changing the world yet. They were about surviving long enough to try.
Here’s what happened next. https://t.co/Vv7c7xiTF1

[OkMan @HeroOkMan]

https://t.co/NkjJFjJsOd

[Arion @ar10nx]

@TeslaXrp Exactly. 1,500 is already enough to expose whether the system can coordinate charging, routing, pickups, and traffic at city scale. Prove that works first. Then 10,000 becomes an execution question, not a spreadsheet fantasy.

[Arion @ar10nx]

Elon Musk: “You don't want to tip over.”

A lunar Starship can shed the flaps and heat shield used for atmospheric flight.

But it needs big landing legs.

In this 2024 tour, Musk explains why: one foot could meet a boulder while another meets a crater.

[Arion @ar10nx]

SECURE THE INGREDIENTS FIRST

Tesla's battery plans reach back to lithium refining and cathode production.

In this 2025 presentation, Musk explains why: a factory needs access to its ingredients through supply disruptions.

Capacity on paper still needs material at the door.

[Arion @ar10nx]

COUNT THE BYTES

Two AI models can have the same parameter count and very different memory requirements.

In this January 2026 discussion, Elon Musk asks for another measurement: the size of the file.

A parameter stored in 4 bits takes less room than one stored in 16 bits. The hardware still has to store those bits, move them and perform the computation.

That makes memory capacity and bandwidth part of the model's practical cost.

Parameter count describes one dimension. The bytes help explain what it takes to run the thing.

[Arion @ar10nx]

An AI benchmark starts with questions humans can understand.

Musk argues that a bigger leap would be questions we struggle even to comprehend.

In this January 2026 discussion, the challenge moves from finding an answer to understanding what is being asked.

[Arion @ar10nx]

Elon Musk: “You can cut your radiator mass in half.”

His February 2026 proposal: raise a space chip's operating temperature by roughly 20% in Kelvin.

He argues that hotter operation could reduce radiator mass.

For orbital compute, tolerating heat can save weight.

[Arion @ar10nx]

Elon Musk's argument begins with what currency really does: it allocates scarce labor and resources.

If AI and robotics can produce most goods and services, that scarcity changes.

Energy does not disappear.

Every computation, factory, vehicle and robot still needs it.

That makes energy the deeper unit beneath the financial layer—the ability to do useful work.

Money can be issued.

Energy must be generated, stored and delivered.

[Arion @ar10nx]

Elon Musk: “Those are the three things that are missing.”

An intelligent mind. A capable hand. A supply chain built for humanoid robots.

Elon Musk's point: the software may be visible, but manufacturing decides the scale.

[Arion @ar10nx]

Elon Musk says the moving conveyor belt is not what makes a production line work.

What matters is linear, adjacent flow: one station leads to the next, every station has a similar work tempo, and labor specializes around each step.

Flow first. Machinery second.

[Arion @ar10nx]

A trillion dollars cannot buy one sandwich on a deserted island.

Elon Musk uses that thought experiment to explain what money really is: a database for allocating labor, materials and services across time.

The number in the account is not the product.

If the goods, skills and energy do not exist, adding more money cannot create them.

That is why financial capacity and productive capacity are not the same thing.

The real bottleneck may be engineers, machines, electricity or time.

Capital can coordinate resources. It cannot substitute for resources that do not exist.

[Arion @ar10nx]

Elon Musk says DOGE was meant to confront America’s deficit and wasteful or fraudulent spending.

“I got a little too involved in politics, got carried away, frankly.”

That was the mission.

Did the political detour help—or hurt—it?

[Arion @ar10nx]

Tesla’s driving AI learned to read road signs without receiving a separate reading lesson.

Elon Musk describes it as an unexpected result of end-to-end training.

The system watched video, connected visual patterns with driving behavior and built a useful representation of written road information.

Text mattered to the task—so the capability emerged.

That is the fascinating and unsettling part of large neural networks.

Train for successful behavior in a rich world, and useful internal skills can appear without being individually programmed.

Photons go in. Driving controls come out. Reading appears somewhere between them.