What we are building
The PeoplesWeb is a project to build a complete technology stack that lets digital platforms run entirely on the smartphones of the people who use them, with no central company in the middle. The same architecture can support the full range of today's platform services: ride-hailing, delivery, freelance and gig work, marketplaces, accommodation, social networks, digital currency and more. The difference is who owns the platform and who keeps the value it creates.
On today's platforms, a single company controls the infrastructure, sets the rules, takes a commission on every transaction, and owns the data. The PeoplesWeb removes that company. Each platform is owned and governed by its own participants — the drivers, the couriers, the freelancers, the members — who connect directly with one another and decide collectively how their platform is run.
Two things about the project have changed what is now possible. The stack is built with AI: rather than being programmed by hand, each part is set out in a scientific paper and AI writes the working code from it, which lets a small academic team attempt a system of this scope. And because a member-owned platform has no operator, it has nothing to hand over to AI — no control over what members see, who counts as a real participant, or what content circulates — at a moment when conventional platforms are ceding exactly that. How both work is set out below.
Why this matters for economic growth
The case for the PeoplesWeb is, first and foremost, an economic one. Digital platforms typically extract around a quarter to a third of the value of every transaction and route it to a small number of very large operators. When that value stays instead with the workers and communities who generate it, more money circulates in local economies, more small enterprises can compete, and the incentive to innovate is restored: when the people who build something useful are the ones who benefit from it, they have every reason to keep improving it.
A common assumption is that worker ownership trades efficiency for fairness. The evidence does not support that. Worker-owned firms have repeatedly been shown to match or exceed the productivity of conventional firms, while delivering greater employment stability and markedly lower internal inequality. They tend to adjust pay through downturns rather than resort to layoffs. The obstacle has never been that cooperatives perform poorly. It is that they have lacked the technological foundation to compete with platform incumbents, and that company founders have had private incentives to choose ownership structures that maximise their own share rather than total value created. The PeoplesWeb addresses the first problem directly, and is designed around an economic model that addresses the second.
An economic model of incentive-compatible cooperatives, developed by Francesco Caselli at LSE — builds market discipline into cooperative structure, so that self-interested members still have strong reasons to work hard, invest, and stay committed. Democratic governance becomes a source of efficiency rather than a drag on it: shared income creates mutual accountability, workers share productivity-improving knowledge because they directly gain from it, and decisions are delegated to elected managers with real authority.
How it works
The PeoplesWeb is built as a layered technology stack. The lower layers — a programming language and runtime for smartphones, and the social-networking infrastructure that lets phones find and communicate with one another directly — turn an ordinary phone into a node in a serverless network. On top of these sit the layers that make a platform a cooperative: a constitutional governance system through which members set their own rules, rates, and budgets and amend them by vote; the economic rules of the incentive-compatible cooperative; and a federation protocol that lets independent cooperatives join together — by sector, by region, or both — into larger structures while each remains in control of its own affairs.
The whole stack is designed to run as a single "super-app" on each member's phone: one app a person installs, inside which each cooperative runs as a mini-app. It is the arrangement people already know from WeChat — except that the host is the person's own device rather than a company's servers, and there is no operator in the middle.
Because there is no central server, there is no single point that can be monitored, charged a fee, or switched off. Each platform runs on infrastructure its members already own. And because the software is open source, anyone can inspect it, run it, or improve it.
The absence of an operator matters for a reason that has become urgent since the project began. On conventional platforms, control is increasingly being handed to AI — over what users see, over who counts as a real participant, and over the content that circulates. A platform with no operator has no such control to hand over. Membership is by personal introduction, so an AI agent can join only when a real person vouches for it and is accountable for it; and every item of content is signed by its author and by each person who forwards it, so the source of a deep-fake is always traceable. Being owned by its members is what makes a platform resilient to AI, not only to corporate capture.
How it is built
The PeoplesWeb is being built in an unusual way, and the method is itself part of the research. The team writes very little of the production code by hand. Instead, each part of the system is first set out in a full scientific paper — the mathematics, the design, the proofs — and AI then writes and tests the code directly from that paper. The paper, not the code, is the source of authority: when AI runs into a gap or an inconsistency, the fix is made in the paper first and the code is regenerated from the corrected version. The result is software whose behaviour is tied, step by step, to a description written for peer review.
This "Human–Science–AI" method is what makes a project of this ambition achievable by a small academic team: the researchers concentrate on the mathematics and the design, and AI does the programming. It is also why the project has moved quickly over the past year.
Where the project stands
The PeoplesWeb is at an early, pre-commercial stage, but it has moved a long way in a short time. The scientific foundations — across distributed computing, the mathematics of democratic decision-making, and cooperative economics — have been developed and published over several years.
Over the past year, the Human–Science–AI method has turned a substantial part of that theory into working code. The project now has its own smartphone programming language and runtime, a type system that catches errors before a program runs, and a first version of the social-networking and super-app infrastructure the rest of the stack sits on. On top of these, several cooperative building blocks have been built and tested as prototypes — including a secure social graph, a social network, a child-safe social network in which parental consent is built into the software itself, and a digital currency with its own financial instruments. Nine such components were carried from paper to tested code in a single year, all released as open source.
Most of these prototypes already run together. A social graph, a social network and a currency now operate as a single app on an ordinary smartphone — an early but real demonstration that the pieces fit together on the device, with no server anywhere.
What remains is to harden these prototypes into software that can match incumbent platforms on scale and security, to build the constitutional governance and federation layers to the same standard, and to bring the whole stack together in real pilots: a portfolio of cooperative apps, tested with real members, and then federated together. That is the goal of the current phase of work.
Who is involved
The Peoplesweb is led by the London School of Economics, where the project draws on the Departments of Mathematics and Economics. The principal investigators are Ehud Shapiro and Francesco Caselli. Shapiro’s work on grassroots distributed systems provides the project’s architectural and mathematical foundation. Caselli is author of the incentive-compatible cooperative model that underpins the project’s economics, and they are joined by Andrew Lewis-Pye (LSE Mathematics) and Idit Keidar (Technion University), who work on the consensus protocols at the heart of the governance layer.
We are working with several partners, including the University of Amsterdam, the European confederation of industrial and service cooperatives (CECOP), the bicycle-delivery cooperative federation CoopCycle, Fondazione Pico, Centro Studio Doc, Camplight and Platform Coops eG.