What’s inside the sphere? The untold story behind Worldcoin

I traveled to Berlin to see the future. Or more accurately, I traveled to Berlin to see the ball – literally – a device that some believe is humanity’s best hope for taming and even harnessing the future power of artificial intelligence. Others believe it’s designed to track and control humans, like something out of a Black Mirror episode.

I am staring at this sphere, the Orb.

The Orb is about the size of a bowling ball. It's alloy-plated, shiny and smooth. I lean in and stare into a black circle, like you would at an optometrist. The Orb then uses an infrared camera, sensors, and an AI-powered neural network system to scan my irises and verify that I am, in fact, a human.

I’m not the first to try this. More than 2 million people have now seen the Orb, the flagship device of Worldcoin , the crypto and AI project co-founded by Sam Altman (CEO of OpenAI ) and Alex Blania (current CEO of parent company Tools for Humanity).

Worldcoin has a bold premise: AI will continue to improve and eventually evolve into AGI (Advanced General Intelligence), meaning it will be smarter than humans, which will spur a leap in productivity, creating wealth that should not be seized by the elite, but rather distributed fairly to all of humanity — literally everyone — in the form of a Universal Basic Income (UBI), which will empower billions of people. UBI will come in the form of Worldcoin.

The merits of UBI have resonated with Altman for years. “Even without talking about AI, UBI is interesting to me, and it’s an idea that appeals to a lot of people, ” Altman said in a recent Zoom interview. “ If our society is wealthy enough to eliminate poverty, then we have a moral obligation to figure out how to do that.”

So, perhaps AI can succeed where political policies have failed?

“In a world of AI, it’s even more important for obvious reasons, and I still expect there will be jobs in a post-AI world,” Altman said. “But, A, I do think we need some kind of buffer in the transition, and, B, part of the excitement about AI is that it’s a world of greater abundance.”

A worldcoin that follows this logic could be the key to providing this material abundance. But there’s a problem. If the goal is to distribute tokens to everyone for free, then in this AI-driven future, how can we be sure that we’re distributing the loot to humans, and not AI-driven dummies? (It’s only a matter of time before AI laughs at CAPTCHAs.) Or what if someone with ulterior motives uses AI to create multiple wallets and cheat the system?

The team thought about this, too, and they explored all the ways they could prove they were human, and came to the painful conclusion that they had no choice . "We really didn't want to do this, we knew it would be painful, it would be expensive, people would think it was weird," Blania said . "But we thought we had to do it: we needed to verify humans with biometric data." "We didn't want to go down this path for a number of reasons, but it was really the only solution," Blania said.

This is the untold story of that solution and the journey to discover whether it is a solution or a problem.

Iris Sphere

The sphere's design is sleek and simple, with no controls or knobs, which is very Apple. This is no coincidence, as the lead designer of the Orb was Thomas Meyerhoffer, Jony Ive's first employee at Apple. (Ive is the legendary designer of the iMac, iPod, and iPhone.) The Orb was designed to be as simple as possible, and Meyerhoffer once said : "It has to be simple enough for all of us. Everybody, everywhere."

At the Berlin office , Blania showed me an old model of the Orb and told me stories about the company’s early days, when they were first tinkering with hardware. The idea was originally conceived as a “Bitcoin project,” with the goal of giving Bitcoins away for free to people once they’ve proven their humanity. Blania held up the old version and laughed because it had a slot that could spit out physical tokens, like a reverse piggy bank, and even had two eyeballs and a mouth.

Early Orbs even talked to people. “This thing talked to you in a robot voice,” Blania recalls. Each early Orb could hold 15 physical tokens (which contained the keys to actual Bitcoins), with the idea that people would take cryptocurrency more seriously if they held something in their hands. (The team quickly abandoned this idea for obvious reasons.) “We tried a lot of things, like having the Orb vibrate when it told people something,” Blania says. They quickly iterated on new versions of the Orb every week using a 3D printer.

Blania, a sturdy, baby-faced 29-year-old in jeans and a T-shirt, has nothing on the CEO look. He’s leading one of the most ambitious projects on the planet, and it’s also his first job. Altman tapped him as CEO and co-founder after working at Caltech, where he researched neural networks and theoretical physics. At first, Blania admits, “I thought I was a terrible CEO, except for my technical depth.”

So Altman got him help. " Altman basically arranged for a weekly meeting with a group of people who would tell me what I was doing wrong," he said. One of the CEO coaches was Matt Mochary, who had previously coached Altman and Brian Armstrong. He taught them the basics of management, such as how to have one-on-one conversations, how to run employee meetings, and how to give public speeches.

Blania doesn’t have many hobbies, except for weightlifting and meditation. He splits his work time between San Francisco and Berlin (Tools for Humanity’s two main offices), where he starts his day at 9 a.m. and leaves the office at 10 p.m. to go to the gym. “I try to work every waking hour,” he says.

The job includes leading 50 full-time employees, some of whom are tasked with building a new crypto wallet from scratch. “The user experience for crypto is terrible,” said Tiago Sada, head of product and engineering, whom I also met at the Berlin office. Sada is another genius. (There are many such people in the AI ​​community.) He grew up in Mexico and built robots with friends at age 14, then founded a startup called “the Venmo of Mexico” before meeting Altman at the Y Combinator incubator.

Sada was initially skeptical of crypto because, in his opinion, it was difficult to get non-technical people to sign up easily. When he asked people to do things like download the MetaMask extension, they were stuck. Sure, crypto wallets work for those who are interested in crypto, but they are a barrier to entry for many of the 8 billion people on the planet. The core idea of ​​WorldCoin is to make cryptocurrency easily accessible to everyone, whether they are tech savvy or not. That is, something they can do in the blink of an eye. So they built WorldApp, which syncs with Orb and allows for near-instant logins.

I tried it myself. In the Berlin office, I downloaded the World App from the App Store. The app synced with The Orb, which sat on the conference room table. A few seconds later, I was gazing at the shiny orb, my account was verified, and I was now the proud owner of 1 Worldcoin*. (*This wouldn’t have been possible if I’d tried it in the US, where they haven’t launched the token for regulatory reasons — at least not yet.)

The sign-up process was pretty smooth, at least for me. If you ignore the whole eyeball scanning thing, it was easily the smoothest crypto onboarding I've experienced in over five years. (More on that later.)

One of the reasons the technology works so smoothly is the embrace of artificial intelligence, whose paradoxes are everywhere. The potential consequences of AI — both wonderful (productivity gains) and malicious (deep fakes) — drive the company’s mission, but on a more day-to-day basis, recent advances in AI are making engineers more efficient. “Without AI, Worldcoin wouldn’t exist,” Sada said . Multiple machine learning models help power Orb, and Sada said the AIs are (actually) already training other AIs, further increasing their productivity.

How does AI train AI?

“People think we need a lot of data to train algorithms, but in reality, many models allow us to generate synthetic data,” Sada said. “Just as you can use Dall-E to create an image of Luke Skywalker dunking in the style of Caravaggio — and it comes back in seconds — engineers can use AI to create data simulations. This allows us to use relatively little data and delete everyone’s [biometric and iris] data by default.”

This brings us to a question that has plagued Worldcoin since its inception: What exactly is Worldcoin doing with these eye scans?

Critics and skeptics slammed Altman after he released The Orb via Twitter in October 2021. Edward Snowden warned in a tweet: "Don't classify eyeballs. Don't use biometrics for anti-fraud. In fact, don't use biometrics for anything."

Snowden acknowledged that the project uses ZK-Proofs to protect privacy, but insisted that "it's clever, but it's still bad, and the human body is not a wicket."

David Z. Morris wrote: "For a private company to collect biometric data on every person on Earth is a huge risk, and by the way, calling the device the Orb - with allusions to the Eye of Sauron - is creepy."

Blania, Sada, and others at Worldcoin have repeatedly told me that the Orb does not collect biometric data from eyeballs, or at least not without explicit permission from the user.

The company’s privacy statement reads: “Privacy is a fundamental human right. Every part of the Worldcoin system has been carefully designed to defend it without compromise. We don’t want to know who you are, just that you are unique,” ​​which means that some data is captured if the user allows it. The default setting is to capture no data; users can change this setting and allow data to be stored, which Worldcoin says is used for the narrow purpose of improving its algorithms. (Why a user would enable this feature is beyond me.)

“The coolest thing, and the hardest thing, is that the Orb handles all the calculations and verification locally to vet whether you are a unique human being, and then it generates a unique ‘iris code,’ ” Sada said. You can think of your World ID as a passport, Sada said, and all the Orb does is stamp your passport to say it’s valid. Again, he stressed, no eyeball image is captured. It’s just a code that says you are a unique person, not your age, race, gender, or eye color.

On the day of Worldcoin’s public launch, Vitalik Buterin wrote a detailed blog post questioning its privacy claims. He raised concerns but gave it a positive review. He said: “Overall, despite the ‘dystopian vibe’ of staring at a sphere and having it scan your eyeballs deeply, specialized hardware systems do seem to do a pretty good job of protecting privacy… However, the flip side is that specialized hardware systems introduce greater centralization issues. As a result, we think cypherpunks seem to be stuck in a dilemma: we have to weigh one deeply held cypherpunk value against another.”

Once users have a WorldID (which Worldcoin insists is privacy-preserving), they can use it as a master key to access other applications and websites (such as Twitter or ChatGPT) in the future. They have already started experimenting with this feature. WorldID recently announced an integration with German identity and access management provider Okta, and more partnerships are in the works.

WorldID is a form of self-sovereign identity, which in itself is the holy grail for many in the Web3 space. In an optimistic and best-case scenario, Orb can scale SSIDs and UBI to billions of people, wresting online “identity” from the hands of centralized corporate giants, and helping poorer, marginalized communities gain more financial empowerment. That’s the vision.

But who pays for this? The way the system is currently configured, once you sign up for Orb, you can claim 1 Worldcoin per week. This is the early kernel of UBI. Who pays for this token that suddenly appears in the wallet (or eyeball) of every person on the planet? On the one hand, there is indeed precedent for cryptocurrencies that came into the world like magic and then eventually increased in value, Bitcoin being the best example, but then again, part of Bitcoin's value proposition is its scarcity, with its supply capped at 21 million.

Token Economics

The token economics of WorldCoin are more ambiguous.

Jesse Walden, an early investor in the project and general partner at Variant, acknowledged that “who pays” is a good question, but said, “I don’t know if there’s a clear answer right now, or if there needs to be one.” In his view, most startups don’t come up with a business model at the beginning, they usually focus on growth, and the growth of the network will eventually generate use cases and value.

Altman has a more pragmatic answer. In the short term, Altman said, “We hope that when people want to buy this token because they believe this is the future, there will be money flowing into this economy. New token buyers are how it effectively gets paid.”

Of course, the longer-term, grander vision is that the fruits of general artificial intelligence will bring economic rewards to humanity. (Hence the name of WorldCoin’s parent company, Tools for Humanity.) How this happens is anyone’s guess. “Ultimately, you can imagine all kinds of things in a post-AGI world, but we don’t have specific plans for that,” Altman said. “At this stage, that’s not the focus.”

Few people are better equipped to envision a post-AI world than Altman, who sits at a strange intersection between the two AI projects. As the most central player in AI development, Altman is a co-founder of a project that aims, in part, to curb the misuse of AI as much as possible. Although he doesn't think so: "I find it more appealing to describe this story as 'create a problem here, solve a problem there'."

This is the model of development as Altman sees it: The world will continue to move forward. As the world continues to move forward, the playing field has changed. There are still a lot of things that should happen, but this doesn’t feel like a solution to the problem. It’s more like a co-evolving ecosystem. I don’t think one is a response to the other.” (I admit that Altman crushes me intellectually . He strikes me as sincere and well-intentioned, but I do find his answers difficult to understand.)

The question I asked Altman was the same one I had asked Blania in our first conversation: What would the world look like if WorldCoin was fully successful and everything worked? Let’s say billions of users joined, and the financial benefits of AGI were distributed fairly to everyone. What would that future look like? What would it all mean?

“I think we’re all going to become the best versions of what we hope to be, with more individual autonomy and agency. More time, more resources to do things,” Altman says. He speaks quickly and without hesitation; these are ideas he’s been thinking about for years: “Just like any technological revolution, people are going to find amazing new things for each other… But this is a very different and more exciting world.”

What are the biggest risks and challenges to achieving this vision? Altman believes that “it’s too early to talk about major challenges,” but he acknowledges that OpenAI and Worldcoin are “a long way from being useful” and that “there’s a lot of work ahead of us.”

The heavy lifting involves launching the Orb, and this is where things get dicey.

Application Scenario

The goal of scanning 8 billion sets of eyeballs is hardly ambitious. In the long run, even without any biometric conditions attached, providing a handful of free “candy” to every person on the planet is a daunting logistical challenge. How to reach remote areas? How to transport the spheres safely? How to explain this complex relationship between the potential of AGI, the need for UBI, and the merits of encryption?

The pitch is essentially, hey, get some free crypto!

The message has been refined, amended, and nuanced over time, but this is the basic pitch: users can sign up for some free crypto, and this Orb is your way of proving you haven’t signed up anywhere before.

Thinking back to Blania’s first field test, it was like a scene from a Judd Apatow comedy. Blania was in a park in Germany, holding the Sphere and looking for possible users, when he noticed two young women. Should he approach them to talk? The Worldcoin team watched from a distance. (In the Berlin office, Blania found a photo someone had taken of the scene on his phone and showed it to me: like a man in a bar, working up the courage to strike up a conversation.)

At that time, they were still in the early days of the "Bitcoin Project". So Blania approached one of the ladies, showed her the Orb, and said it could help her get free Bitcoin. He told her: "The only thing this device does is to make sure you only get it [Bitcoin] once, but you get Bitcoin, and that's something to be happy about."

The woman’s response was simple: “Are you crazy?”

She chose not to have her iris scanned, but her friend did.

"Actually, I think she just thinks I'm cute," Blania said with a half smile.

This isn’t surprising. Blania is a good-looking guy, smart, and able to clearly and convincingly speak about the benefits, nuances, and raison d’être of WorldCoin (who wouldn’t be impressed?). But how does Alex Blania scale? Maybe if Blania could be cloned, he could personally sign up 8 billion people.

But back to reality, in the beginning, Blania and a small team just dragged the sphere around the streets of Berlin, showing it to people and trying to explain it on the fly. "That was actually in the early scripts, people would approach us because it was this shiny sphere, and people would say, 'What the hell is that?'" he said.

Early orbs spoke to users in a strange robotic voice, instructing them to move closer or farther away, or perhaps to the left. (The team later made a series of improvements to automate this process.) The robotic voices confused onlookers, who sometimes took hilarious selfies with the Orb .

The neo-colonial problem

Less interesting are early attempts to recruit users in Nairobi, Sudan, and Indonesia. In April 2022, MIT Technology Report published a 7,000-word feature titled “Deception, worker exploitation, and cash handouts: How WorldCoin recruited its first 500,000 test users.” The authors argue that while the project is ambitious, “so far all it has done is build a biometric database based on the bodies of poor people.”

The report describes a shoddy operation riddled with misinformation, data gaffes, and malfunctioning orbs. “Our investigation revealed a large gap between WorldCoin’s privacy-focused public messaging and its user experience, and we found that company representatives used deceptive marketing tactics, collected more personal data than it acknowledged, and failed to provide meaningful access to information,” Eileen Guo and Adi Renaldi wrote.

I mentioned the report to Blania and Altman . “The first thing to understand is that this article was published before the company had completed its Series A,” Blania said. He acknowledged that this was no excuse, but stressed that the project was in its infancy, and that “virtually everything has changed” since then, with operations and protocols becoming more stringent. Of course, it’s the possibility of errors like this — no matter how good the team’s intentions — that makes people uneasy about sharing their biometric data.

Blania was also outraged that the article was framed as (in his words) "a colonialist attempt to get poor people around the world to sign up". He said this was misleading, as more than 50% of the sign-ups at the time were from wealthier countries such as Norway and Finland and European countries. The goal was to test sign-ups in developed and developing regions, hot and cold climates, urban and rural areas, to better understand what works and what doesn't.

Altman believes that missteps are a natural part of any large-scale project. "With any new system, you're going to face some initial fraud, and that's part of the reason we're doing this for a long time in [this] slow testing phase. Learn how the system faces abuse and how we're going to mitigate that. I don't know of any system that's this scale and this ambitious that doesn't have a fraud problem. We want to be very thoughtful about this."

One of the mitigation measures was to change the way Orb operators are compensated. There are currently 200 to 250 active Orbs in the space, and about a few dozen operators, each of which employs their own sub-teams. In the beginning, Worldcoin simply paid operators based on the number of raw registrations, which led to some sloppy and shoddy practices.

Blania says operators are now incentivized not just by the number of signups, but also by the quality of signups and how aware users are of what’s going on; after a few weeks of Orb scanning, if users actually use Worldapp, operators will get paid more. (The main way you “use” Worldapp right now is by claiming the weekly Worldcoin.) Two operators in Spain I spoke with, Gonzalo Recio and Juan Chacon, largely endorsed the new protocol, but whether the process is followed closely around the world remains an open question.

How can people trust that Worldcoin actually solves these problems? Altman hears these questions, and he knows he’s unlikely to win over the skeptics, but he doesn’t seem to mind. He believes the more convincing answers will come not from him, Blania, or the company, but from Worldcoin’s early users. “You can try to answer a bunch of questions and do all these things, but that’s not really how it works,” he said.

“What really works is the first million people, the early adopters, the ones who move forward — convince the next 10 million. And then the next 10 million become closer to the norm. They convince the next 100 million. And these are really the norms that convince the other billions.”

Policy and the future

In our first conversation, Blania told me that if WorldID and Worldcoin’s UBI were to be adopted across the board at scale, it would “probably be one of the most profound technological changes ever.” If that were indeed the case, would this create a new set of complex legal, policy, and even existential questions that governments would need to consider?

As I wrapped up the meeting in our Berlin office, one final question stuck with me. It almost felt like the project was so ambitious, so wild, so transformative — at least in theory — that the powers that be hadn’t fully considered the consequences. If humans were compensated not through the sweat of their labor but through general artificial intelligence, wouldn’t that represent a fundamental shift in the way the world is structured? Wouldn’t governments insist on regulating this? Assuming the answer is yes, how would WorldCoin solve this problem?

Blania leaned back in his chair and thought for a moment. He said, "This is obviously a major discussion point, and I'm going to start with the most important thing right now. And the most important idea is actually far less complicated than all the things you just mentioned." Blania said that they were focusing on just the basic points of regulatory uncertainty in the United States, and that Worldcoin "may well be the largest cryptocurrency transaction ever in the world."

Altman pushes back against the idea that policymakers are clueless or are just scratching their heads. “I’ve been to about 22 countries and met with a lot of world leaders, and people understand this better than I thought and are taking it very seriously,” he says. (It’s unclear whether by “it” he’s referring to the specific ambitions of WorldCoin, or the broader challenges posed by AI.) “I’m spending more and more of my time now, not on the technology issues, but on the policy challenges. Ultimately, in order for the world to get through this in a good way, it’s going to have to be a two-part solution, a technology and a policy one. The policy part is probably going to be more difficult in some senses.”

Policy hurdles are one risk for Worldcoin. Not scaling fast enough is a risk for Worldcoin. More missteps in the field, like the ones highlighted in the MIT technical report, present another risk. Or iris data could be leaked. Or token economics fail. Or an erosion of trust that discourages future signups. Or the logistical challenges of bringing the Orb to trickier corners of the world. Or technical and manufacturing glitches. Or the discovery that the Orb has been compromised in some way. (As Vitalik Buterin points out, “The Orb is a hardware device, and we have no way of verifying that it was built correctly and has no backdoors.”)

Or maybe WorldCoin will never be worth more than a few cents, so no one cares. There are many, many risks on the long road to widespread adoption of WorldCoin, and the project is still a moonshot.

But as the company’s think tank sees it, the biggest risk isn’t the risk to WorldCoin itself. The most chilling risk, to them, is that something like a biometric WorldID will be developed, just not in an open or privacy-preserving way.

Altman’s original logic, articulated before ChatGPT went mainstream, remains compelling: eventually AI will get so good that it can easily pretend to be human, so we’ll need a way to prove our humanity. Perhaps biometric proof is inevitable. Who should be the stewards of that solution? “Something like WorldID needs to happen,” Blania says. “You need to authenticate yourself online. Things like that are going to happen. The default path is that it’s not intrinsically online, it’s not privacy-preserving, and it’s fragmented by governments and nations.”

For Blania , the non-private version of biometric vetting is the real Black Mirror plot. “This is the default path, and I think Worldcoin is the only way,” he said .