Skip to content

The CCS Interview

Perceptron's Peter Anthony: 800,000 nodes challenge the AI data monopoly

Peter Anthony, co-founder of Perceptron Network, explains how users earn by running nodes and completing data tasks, and why he thinks decentralized data collection can undercut centralized AI data providers. The conversation was recorded in July 2026, when the project was still pre-TGE and users were earning points. At the time, Anthony said he expected a token launch within eight to 10 weeks, depending on exchanges.

Interview by Ashton Addison Recorded July 2026 5 min read

Key takeaways

  • Perceptron operates a decentralized network of 800,000 nodes that gather AI training data at 92% lower cost than centralized providers like Scale AI.
  • Users earn points by passively running nodes to share bandwidth or actively annotating data and contributing to specific datasets.
  • The platform acquired Router, a crypto company that automates user contribution tracking and reward distribution across the network.
  • Perceptron sells gathered data to AI model builders and provides live data sources for sentiment analysis to exchanges and quant desks.
  • Rare language contributions command higher rewards than common tasks, with the Data Questing Platform determining compensation based on task difficulty and scarcity.

How Perceptron's node network works

Ashton Addison: Big tech sells packaged user data for millions while users see nothing. Is that what Perceptron is built to fix?▶ 02:37

Peter Anthony: At the moment, when you and I contribute data that Google or others use, they package it up and sell it on for millions, if not hundreds of millions, of dollars. Perceptron lets an individual user run a node on a computer or an Android device. You're sharing your excess bandwidth, which gives us a vantage point on the internet. We're not gathering your data. We look through your access point to the internet, your IP address, to gather a package of data. We have around 800,000 downloaded nodes, and each one gives us a vantage point from which we gather pieces of data and build them into meaningful data sets. By running a node you earn points, which will translate into tokens in future. Raw data isn't always as valuable as it can be. Think of milk: it's cheap, but cheese and ice cream are the same product made more valuable. Annotating data does the same. Users can also take part in annotation tasks and contribute to data sets, for example by taking photos of certain places in certain cities, and earn rewards for those tasks.

Ashton: So there's a passive side and an active side, and the active side boosts your earnings?▶ 05:19

Peter: Like anything, the more difficult something is to deliver, the higher its value. But there is a passive contribution people can make by running nodes. What we do is take data and sell it on to clients who are building things like AI models. We also gather live data sources for sentiment analysis, which is particularly useful in crypto and can be sold to exchanges and quant desks. That creates revenue, which is distributed among contributors. It also goes into developing future products and onboarding more clients.

What is very difficult is building a network of hundreds of thousands of people.

Earning rewards and keeping data quality high

Ashton: How is the amount a contributor can earn determined?▶ 06:54

Peter: As we move towards completing our data questing app, some data sets will be much easier to contribute to. If we need something read out in English, lots of people can do it, so it's a lower-value task. If we need a rare language from sub-Saharan Africa and only five or 10 users on our network can read that passage, that would command a higher reward. The mechanism for gathering points is complex and heavily regulated by AI. We've just acquired a company that built a framework to identify each user and how they've contributed to the network. So it's done automatically, and nobody is sitting down working out individual contributions.

Ashton: What stops people from rushing through tasks and submitting junk just to get paid?▶ 14:00

Peter: We've been looking at this for the data questing platform, to identify serious versus non-serious contributors. People who try to do every task without doing it properly will face a penalty. Say someone requests 500,000 images of cats. We build an understanding through AI of what a cat looks like, the AI trains on that, and then it gathers large inventories of images. When those come back, the AI checks them against criteria for the annotation, and anything not in line with what it expects is rejected. That happens for one of two reasons. Either it wasn't done properly but you can see an effort was made, which is something we need to work out. Or it was done with no intent to do it correctly, in which case penalties should apply. People shouldn't fear contributing because they might get it wrong, as long as they're making an effort. But there should be penalties in place so people do it properly.

Breaking the data monopoly

Ashton: How do your economics compare with centralized AI data collectors, who reportedly paid around $200 million for Reddit data?▶ 10:10

Peter: One of the major problems in AI is the monopolization of access to data and compute, predominantly because of cost. Traditionally, data has been gathered through APIs for Twitter, Reddit and so on. That data is extremely valuable, so if you own those companies you can charge huge amounts for it. Then only OpenAI, Anthropic and a handful of others can afford it. If you're a new AI project, you'll never access that quality of data, so your product won't be very good. You could be the smartest person ever, but if you've never read a book, you won't know much. What we do differently is look through the eyes of an individual user, hundreds of thousands of times, to gather what would be shown through APIs. We access only what any ordinary user can reach without going through paywalls. If you can access Twitter or Reddit from where you are, I can too, hundreds of thousands of times over, and gather the same meaningful data sets for a fraction of the cost. I can gather the data effectively for free, although we still need money to build and run the network and distribute to users. So we're breaking down the monopoly by being decentralized.

You could be the smartest person ever, but if you've never read a book, you won't know much.

The data fund, business intelligence and the token launch

Ashton: Why launch a $10 million AI data fund, and how does it tie into the network?▶ 16:43

Peter: I've been in crypto a long time. I founded the House of Crypto YouTube channel and have watched a lot of projects come and go. I like blockchain and what it can offer, but it's rare to find a case where it's an essential component that improves how a service or business works. With AI, there are several ways in which blockchain is the quintessential technology to push things forward. I wanted to showcase what it can achieve by solving a real issue, and what better way than giving AI projects the chance to use our quality data? The aim is to shine a light on blockchain and show that decentralization can work far better than centralization. It's also to enable AI projects to build into the future, since they'll be future paying customers, and to give something back to the community.

Ashton: Where do you want Perceptron to be in the next 12 to 24 months?▶ 20:22

Peter: Our main goal is to build out the data questing platform and move towards live business intelligence. Existing crypto companies already gather and sell data, such as tweets or images for training text-to-image models. That's fine, but I think decentralization and huge numbers of vantage points become really valuable for business intelligence models. Where you look at the internet from differs from where I look at it from. E-commerce companies want to know where their products rank on Amazon today in New York versus Dubai versus Brazil. That's very difficult for a centralized model to see through all those vantage points, and it's where our moat comes in. The tech for a node isn't particularly difficult. What is very difficult is building a network of hundreds of thousands of people. I learned that building my YouTube channel to 300,000 followers, which took five or six years. Centralized companies are not able to do that. We're building towards products for e-commerce, comparison sites and crypto sentiment analysis, sold to exchanges and trading desks.

Ashton: Is there a timeline for the token launch, when points convert to tokens?▶ 23:45

Peter: We're moving towards TGE. Everybody always wants to know when. It's the same in crypto: everyone wants to know the best time to buy, which is actually the most difficult question. We see a TGE coming very soon. It's hard to give exact dates because that hinges on exchanges and how long their due diligence takes, so I'm not going to promise anything. We're looking at the next eight to 10 weeks, to be generous. It could be much sooner than that.

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

Peter Anthony
Peter AnthonyPeter Anthony is the co-founder and chief executive officer of Perceptron Network, a decentralized AI data network. He is based in the United Kingdom and began working in crypto around 2019 after a background in trading forex, stocks and traditional markets. Around 2020 he started… Full profile →
Share this interviewXLinkedIn
▶ Watching the interview