Skip to content

The CCS Interview

World's DC Builder: AI agents need proof of a human behind them

DC Builder, a research engineer at the World Foundation, has worked on World ID and Tools for Humanity since 2022. In this June 2026 conversation he explains how the Orb proves humanity privately, why World built AgentKit with Coinbase, and what he expected for agent payments at the time: far more normal within about a year.

Interview by Ashton Addison Recorded June 2026 6 min read

Key takeaways

  • World ID uses custom hardware orbs with cryptographically signed camera feeds to verify humans at scale while preserving privacy through zero-knowledge proofs.
  • Standard smartphone biometrics like Face ID would create collisions at 15-30 million users, requiring custom imaging hardware for billions-scale uniqueness verification.
  • AgentKit integration with Coinbase links AI agents to verified human identity without exposing personal data, enabling secure agentic commerce.
  • The internet lacks a native identity layer similar to how it lacked native payments before cryptocurrency, making proof-of-humanity infrastructure critical.
  • World ID operates orbs across approximately 20 countries including Portugal, Argentina, Thailand, UK, and Germany to increase accessibility.

Why the internet needs proof of humanity

Ashton Addison: What are World ID and Tools for Humanity, and how does that tie into the growth of AI agents?▶ 00:41

DC Builder: I joined in July 2022, before ChatGPT launched. I was interested in AI as a high school student, around 2019, before I came into crypto. World was co-founded by Sam Altman, Alex Blania and the rest of the team because they could extrapolate the pace of progress and predict what the world would look like. It converges on a world where you can't tell AI-generated content from human-generated content, whether that's text, video or images. World set out to build a way to prove you're interacting with real humans on the internet and on blockchains. Even on Zoom, I want to know I'm talking to the real Ashton and not an AI-generated deepfake. It also embodies a lot of crypto's technologies and philosophies: privacy, decentralization, permissionlessness and scalability.

Ashton: AI can now fake identity, video and images. Has that made the original problem even bigger?▶ 03:12

DC: Yes, and it's going to become ever more prevalent. Right now you can still tell if a video is AI generated. Give it one or two more years and it will truly look indistinguishable. The internet also doesn't have a native identity layer, just as it didn't have a native payments layer. We're only now building crypto tools ready for mass adoption, with stablecoin payments for the internet. Agents can now use the internet on behalf of humans, at a much faster rate and across many applications at once, like booking travel or paying for orders and groceries. There needs to be a way native to the internet to prove you're a real human, and I think crypto and World ID play a core role in that.

There needs to be a way native to the internet to prove you're a real human.

The Orb and the privacy trade-off

Ashton: How does World ID prove someone is human? Do you upload an ID or do a live video?▶ 04:40

DC: We looked at the existing solutions, like passports and biometric-based systems, and none were scalable, private and decentralized at the same time. So we built our own, the Orb. It proves you're a real and unique human in a fully privacy-preserving way, and no information leaves the device. After you verify in person, you have a verified World ID. Using zero-knowledge cryptography, you can then create a proof of humanity that anyone else can verify. An application can check that proof and learn nothing about you except that you're a real and unique human.

Ashton: Why is physical hardware necessary, rather than using your phone as people do for exchange KYC?▶ 06:13

DC: KYC on exchanges usually means taking a picture of someone's ID and storing it. That violates privacy, because you have to keep a database of credentials. To measure that someone is real and unique at scale, the Orb works as a trusted camera that you can't easily spoof. The camera feed is cryptographically signed, and the machine learning models that check the person is real and unique run locally, so nothing leaves the Orb. Measuring uniqueness across billions of people also needs a resolution that phone cameras don't have. If you used Face ID this way with a regular iPhone camera, you'd get collisions at around 15 to 30 million users. Building our own hardware was a painful conclusion, because of production, assembly and distribution, and you can't use the network effects of existing devices like phones.

Ashton: How do people actually get access to an Orb, and how widely available are they?▶ 08:08

DC: There's a website, world.org/find-orb, where you can find Orbs in all the markets. I think it's roughly 20 countries at the moment. Off the top of my head, that includes Portugal, Argentina, Thailand, the UK and Germany. The plan is to make Orbs available in as many places as possible. We've been operating them since 2021, and I think about 18 million users have verified, the last time I checked. We're also creating other proofs of human, such as KYC credentials where zero-knowledge proofs let you prove you're over 18 without revealing who you are. That isn't as robust as the Orb, because you can have multiple passports, and it wasn't designed to solve proof of human.

Real humans on Tinder and Zoom

Ashton: What are the use cases beyond crypto, such as the Tinder and Zoom partnerships?▶ 09:39

DC: The main mainstream areas are gaming, dating and social media. In multiplayer and ranked games, you don't want to play against bots. On Tinder, you could be talking to an AI-generated image of someone pretending to be who they're not, and that breaks what you're there for. Our partnerships with Tinder and Zoom are about enabling the real human experiences people expect. On Zoom, a World ID verified badge could show that both of us are verified and that the Zoom feed matches the image the Orb took, which is stored on the user's device. You create a zero-knowledge proof on your own device, Zoom verifies it, and it can show this is the same human who went to the Orb.

Give it one or two more years and it will truly look indistinguishable.

AgentKit and the race for agent payments

Ashton: One person can run thousands of AI agents. How does World handle agents acting on a human's behalf?▶ 12:09

DC: Imagine you're a music artist selling tickets on Ticketmaster. Scalping is where a bot, which doesn't even have to be AI, buys thousands of tickets and resells them at a markup, so real fans can't buy. These experiences will get worse as AI improves, because captchas can now be trivially automated. AgentKit is the tool we built to solve this. You can register an agent in AgentBook within AgentKit, so you know with certainty that the agent is acting on behalf of a real human verified with World ID. As an artist, you could restrict sales to, say, five tickets per human, so the tickets go to real fans rather than bots.

Ashton: Bot owners have no incentive to get a World ID. Do you have to force partners like Ticketmaster to enforce it?▶ 14:49

DC: It's not about forcing. It's about creating a value proposition that's appealing to businesses. If their customer experience is degrading because of bots and AI, and it will get worse, there need to be multiple solutions, and we're trying to offer the best one. We work with partners to build native tools for their platforms. We partnered with Coinbase because they built x402, an agentic payment standard. Businesses can take stablecoins and associate a World ID proof with their x402 payment, and that's how AgentKit is built under the hood. We've also learned a lot from partners, including things we didn't fully grasp until we kept iterating on our products.

Ashton: Are people actually using agents to pay with stablecoins yet?▶ 16:43

DC: It's very new, in the early adopter phase. People who are technically able or crypto-native are starting to use it, but most experiences are still card or bank payments. Stripe, which I think processes something like 2% of the world economy, has products like Tempo that do the same thing, with an equivalent to x402 called the machine payments protocol, MPP. So even a big business like Stripe is moving into agentic commerce. It's a race. Eventually banks may run on crypto rails instead of old COBOL or Swift. Give it two to three years and I think most payments on the internet will run on these rails, maybe without people knowing. You'll just see 'pay with my dollar wallet,' and it's their bank using agentic payment rails.

Ashton: How soon could we see a big jump in everyday use of AI agents and AgentKit?▶ 21:29

DC: Within a year, a lot of this will be much more normal than it is now. Models are getting really smart. I mostly interact with them as a developer, and Codex is my daily driver. The customer-facing products from labs like Anthropic, OpenAI and Google will take time to reach the same quality and polish as products not driven by agents. Give it 12 months and you'll see a lot more of this. For it to be the prevailing way of paying and using applications, it's probably two to three years. I rarely code by hand anymore. I just talk to a chat box. As the tools improve, agents will use the browser for me, and I'll see things pop up on my devices to approve or change.

This interview has been edited for length and clarity. Recorded June 2026.

DC Builder
DC BuilderDC Builder, who uses the handle dcbuilder.eth, is a Research Engineer at the World Foundation, where he works on proof of humanity and digital identity infrastructure. He has worked on the World ID protocol since 2022, first at Tools for Humanity and, since February 2024,… Full profile →
Share this interviewXLinkedIn
▶ Watching the interview