The Ability to Look Stupid Again

There is something I find genuinely fascinating about Mark Zuckerberg, and it has almost nothing to do with how much money Meta is willing to lose.

It is his willingness to repeatedly look stupid in public after already becoming one of the most successful founders in history.

Seventy-seven billion in losses #

The financial losses are enormous. Meta's Reality Labs lost $10.2 billion in 2021 and $13.7 billion in 2022, according to Meta's own financial results. It then lost another $16.1 billion in 2023 and $17.7 billion in 2024. In 2025, the loss increased again to $19.2 billion. That is roughly $77 billion in Reality Labs operating losses from 2021 through 2025 alone.

But the money is almost the least interesting part of this.

Meta can afford to lose money. Zuckerberg can afford to lose money. The company generated more than $115 billion in operating cash flow in 2025 alone. When you control a machine that large, funding years of expensive technological experimentation is unusual in scale, but not difficult to explain.

What is much harder to explain is why Zuckerberg keeps attaching himself to the experiments.

Renaming the empire #

He could have funded Reality Labs quietly. He could have put an executive in charge of presenting the strange prototypes. He could have waited for products to become polished before appearing next to them. If something failed, it could have been a failed Meta project rather than another failed Mark Zuckerberg vision.

Instead, in 2021 he renamed Facebook itself to Meta.

This was not some side project hidden inside an innovation lab. Zuckerberg published a founder's letter saying that "from now on, we will be metaverse-first, not Facebook-first" and that the metaverse would "touch every product we build." (Read Zuckerberg's original 2021 founder's letter)

He took one of the most recognizable technology companies in the world and effectively attached its identity, and his own, to a technological thesis that was nowhere near proven.

Protection from being fired, not from being laughed at #

There is an obvious objection here: Zuckerberg can do this precisely because he is unusually protected.

That is true.

Meta's own filings say Zuckerberg controls a majority of the voting power of the company's outstanding shares. That gives him control over matters including the election of directors and major strategic decisions. Unlike most public-company CEOs, he does not face the same possibility that a few disastrous quarters or an unpopular long-term investment will simply cost him his job.

But voting control protects his position. It does not protect his reputation.

A dual-class share structure can stop a board from removing you. It cannot stop millions of people from deciding you have lost your mind.

And that distinction is the entire thing I find interesting.

In 2022, Zuckerberg posted an image celebrating Horizon Worlds launching in France and Spain. It showed his cartoon avatar standing in front of primitive-looking versions of the Eiffel Tower and Sagrada Família. The internet absolutely destroyed it. The image became a symbol of everything people thought was ridiculous about Meta spending billions of dollars trying to build the metaverse.

Zuckerberg responded by acknowledging that the graphics were basic and posting previews of improved visuals.

Imagine the incentives at that point.

You have already built Facebook. You own Instagram and WhatsApp. You are one of the wealthiest and most recognizable founders alive. You do not need another successful company to validate your career.

And millions of people are now looking at a cartoon version of your face next to a low-resolution Eiffel Tower and asking why you are spending billions of dollars on this.

The easiest thing in the world would be to become less visible.

Instead, Zuckerberg kept putting himself on stage.

A rocket exploding looks heroic #

That difference becomes especially obvious when you compare him with someone like Elon Musk, who also has an extraordinary tolerance for public failure.

SpaceX has repeatedly blown up enormous rockets in front of the entire world. The first integrated Starship test flight in April 2023 exploded minutes after liftoff. Since then, several Starship tests have ended in explosions or other major failures as SpaceX continued its rapid flight-testing approach. Reuters has described SpaceX's engineering culture as unusually risk-tolerant, deliberately pushing hardware toward failure and iterating through repeated tests.

But there is an important psychological difference between those failures and Zuckerberg's.

A rocket exploding looks heroic. A low-resolution Eiffel Tower looks lame.

When a 120-metre rocket leaves the launchpad, flies through the atmosphere and explodes, everyone understands that you are attempting something extraordinarily difficult. The failure itself looks dramatic. There is a giant fireball reminding everyone that this is rocket science.

Zuckerberg has repeatedly endured the opposite kind of failure.

Sometimes the product does not look ambitious. It just looks bad.

The avatar looks ridiculous. The virtual world looks empty. The headset is awkward. The glasses malfunction during a live demo. The supposedly futuristic experience looks worse than technology people were using years earlier.

There is no giant explosion to make the failure look cool.

You just look stupid. And then you have to come back on stage again.

Live on stage, nothing works #

That is why the Meta Connect 2025 demo was such a good example. Zuckerberg was demonstrating Meta's latest smart-glasses technology when multiple live demos broke. During one demo, the glasses failed while a creator tried to use Meta AI to follow a recipe. During another, Zuckerberg could not answer a WhatsApp video call through the glasses.

Meta CTO Andrew Bosworth later explained that the first failure happened because the demo inadvertently activated Live AI across a huge number of glasses in the building and overloaded the development server. Meta had effectively DDoS'd itself. The second failure came from a race-condition bug Meta said it had never encountered before. (TechCrunch's account of the failed Meta Connect demos)

Zuckerberg's response on stage was basically every founder's nightmare: they had practiced the demos repeatedly, and then they broke when everyone was watching.

At his level of success, none of this exposure is necessary.

That is what makes it interesting.

Reputation becomes the scarce thing #

Once someone becomes extremely successful, reputation becomes one of the few things left that still feels scarce. They already have money. They have access. They have employees. They have distribution. They have people willing to take their calls.

What they can still lose is the perception that they know what they are doing.

That creates a strange incentive where you can become so successful that you stop doing things which might make you look unsuccessful.

You announce products only once they are polished. You let other executives present experiments. You stop making aggressive predictions. You avoid prototypes that might break. You avoid ideas that sound ridiculous before they become obvious.

You stop putting yourself in situations where some teenager on Twitter can quote-tweet your keynote with, "bro spent $20 billion on this 💀."

Eventually, you can become a curator of your previous success rather than a builder of the next thing.

Zuckerberg seems unusually resistant to that.

And importantly, the answer has not simply been to stubbornly keep building the exact same metaverse he presented in 2021.

The bet keeps changing shape #

The implementation has changed enormously.

The original vision placed enormous emphasis on immersive virtual worlds and VR. Over time, Meta pushed further into mixed reality, then camera-equipped smart glasses, then AI glasses, then glasses with displays, while continuing longer-term work on true augmented reality.

Orion is probably the clearest expression of where that journey has gone. Meta says the AR glasses were nearly a decade in the making. The current Orion device is not being sold to consumers; Meta has instead given prototypes to employees and selected external testers while it works toward eventual consumer hardware. (Meta's Orion announcement)

The form of the bet keeps changing, but the underlying thesis is surprisingly consistent: Zuckerberg believes computing eventually moves beyond spending our lives looking down at rectangular screens, and Meta wants to own a major part of whatever comes next.

Some pieces of that strategy have also started producing real consumer traction.

The Ray-Ban Meta glasses are the obvious example. EssilorLuxottica said in early 2025 that more than two million pairs had been sold since the second-generation glasses launched in September 2023, and the company was preparing manufacturing capacity for as many as ten million smart glasses annually by the end of 2026.

That does not retroactively make every metaverse decision correct.

Horizon Worlds did not become the universal social environment implied by the excitement of 2021. VR did not replace smartphones. Billions were spent on products and initiatives that did not achieve anything close to the scale of Meta's existing businesses.

The useful version of stubbornness #

And there is an important survivorship problem here too. Willingness to spend your reputation is not automatically a virtue. Plenty of founders repeatedly make public bets, refuse to retreat, and eventually discover that everyone laughing at them was simply right. Persistence does not turn a bad thesis into a good one.

The useful trait is not an inability to feel embarrassment, and it is definitely not refusing to change your mind.

It is being willing to absorb embarrassment without allowing embarrassment itself to become the reason you stop experimenting.

That seems to be the distinction Zuckerberg has managed surprisingly well. The implementation changes. Products get killed or repositioned. AI suddenly becomes central. Smart glasses become more important than virtual avatars. The technology evolves. But a failed version of the idea does not automatically end the broader attempt.

There is a strange irony in this.

Starting something when nobody knows who you are feels terrifying, but reputationally it is actually cheap. Your first terrible product is seen by almost nobody. Your failed startup becomes a line on your LinkedIn profile. Your stupid prediction disappears because nobody was paying attention when you made it.

Success makes experimentation more expensive.

Once millions of people know your name, every bad demo is clipped. Every failed prediction gets resurfaced. Every ugly prototype becomes a meme. Every pivot can be framed as proof that the previous vision failed.

At that point, protecting your reputation can start feeling rational.

You have spent decades accumulating status, and now every ambitious experiment puts a tiny piece of it back on the table.

Zuckerberg could protect his indefinitely. His voting control protects his job, Meta's cash flows fund the experiments, and his existing accomplishments are already enough to secure his place in technology history.

Yet he keeps standing next to unfinished things.

That, more than the billions Meta spends, is what I find impressive.

A lot of founders are comfortable risking their reputation when they do not have one yet. Far fewer seem willing to build a massive reputation, become one of the most successful people in their industry, and then voluntarily put themselves back in a position where the entire internet might look at what they are building and say:

What the hell is this?

And then ship the next version anyway.

---

Images: cover by Anthony Quintano (CC BY 2.0); 2025 portrait by Jeff Sainlar for Meta (CC BY-SA 4.0); e-foil frame from Zuckerberg's July 4, 2021 Instagram reel (CC BY-SA 4.0); White House dinner photo public domain; Ray-Ban Meta photo by CCadio (CC BY 4.0).