The Socialisation of Platforms

10.09.2026 · 11 min read

Yanis Varoufakis has a name for the world of apps we live in: not capitalism exactly anymore, but technofeudalism — a handful of firms enclosing "fiefdoms" that the rest of us rent access to, on their terms. His answer, when asked what to do about it, is the line that's been rattling around in my head ever since: "not the nationalisation — because I have antipathy towards the state — but the socialisation of apps, of algorithms." Concretely: what if your posts, your followers, your messages weren't trapped inside one company's walls, but simply propagated across the whole network — the way email already works?

That word — socialisation — is doing more work than it looks like. It's worth being precise about what it actually means, because it turns out there isn't one mechanism, there are three, and they show up across every example in this post.

Three ways to socialise a platform

The first is ownership: the people who do the work, or use the service, own the company outright. This is the classic cooperative model — a worker- or user-owned firm instead of a VC-backed one.

The second is protocol: nobody owns the network at all. Instead there's a shared, open standard, and anyone can build a competing app on top of it — the way any mail provider can talk to any other.

The third is governance: a small, deliberately non-extractive organisation — usually a nonprofit — maintains the shared infrastructure everyone depends on, funds itself modestly (often through an optional paid convenience layer on top of a free core), and stays open to outside contribution so the whole thing doesn't stall.

Once you see the three of these as separate moves, the examples stop reading as a list of activist experiments and start reading as a pattern with real range.

Owning the platform

The clearest test case is The Drivers Cooperative in New York. Launched in 2021, it grew fast — over 9,000 drivers, 165,000 trips, $6.1 million in revenue by 2022 — and then contracted hard, down to fewer than 50 active drivers by the summer of 2024. The postmortem calls this the "Blown Head Gasket Effect": running its own tech stack cost the co-op roughly $600,000 a year, with engineers paid $150–200 an hour against drivers earning $20–40. The thing that nearly killed the most prominent worker-owned rideshare in the US wasn't a lack of demand, and it wasn't a lack of belief in the idea. It was the plain, ordinary cost of building and maintaining software. I'll come back to why that matters.

It's not the only model, and it's not always this fragile. Green Taxi Cooperative in Denver organised as a unionised cooperative from the start, backed by CWA Local 7777, and within a year had recruited about a third of the city's taxi drivers — proof that cooperative ownership and organised labour aren't competing ideas, they're the same instinct wearing two hats. Fare.coop runs a hybrid version in the US, splitting ownership between drivers, founders, and a tech company. And in February 2026, the Indian government launched Bharat Taxi — the country's first cooperative ride-hailing platform, explicitly positioned against Ola and Uber — which is a genuinely different route to the same outcome: socialisation from the top down rather than the bottom up.

The strongest existing proof of concept, though, might be CoopCycle — a federation of independent, local bike-delivery cooperatives across France, Spain, Germany, and Denmark, all running on one shared, white-label delivery platform, licensed under a "Coopyleft" that legally blocks any VC-backed company from forking it. Because the engineering cost is shared across the whole federation instead of rebuilt by every city, member co-ops pay roughly 2% of revenue instead of the cut Deliveroo or Uber Eats would take. Fairbnb.coop does something similar for short-term rentals, redirecting half its fee into local community projects instead of shareholder returns. And Stocksy United — a contributor-owned stock photography cooperative running profitably since 2013 — is the reminder that this isn't only a plausible idea; it can last a decade.

The protocol route

Ownership isn't the only way to take a network out of one company's hands. Sometimes nobody owns it at all.

This idea has its own paper trail. In 2019, Mike Masnick wrote an essay called "Protocols, Not Platforms", arguing that the early internet ran on open protocols — email, Usenet — where no single company controlled the network and switching providers didn't mean losing your identity or your contacts. Centralised platforms replaced that with single-company control, which is exactly why moderation fights, lock-in, and algorithmic grievances feel so intractable now: they're structural, not incidental. His proposed fix is precisely Varoufakis's "socialisation of algorithms" idea, just named seven years earlier — rebuild social media the way email works.

It's actually happening, unevenly. ActivityPub is the working version: Mastodon, Pixelfed, Lemmy, and PeerTube are all independent apps that federate over the same open standard, so an account on one can follow, reply to, and be followed by accounts on any other. Meta's Threads has partially joined this since 2023, letting opted-in users be followed from Mastodon — but it's a genuinely partial interoperability. Direct messages don't bridge across. Content warnings don't display. A meaningful chunk of the Mastodon community has simply refused to federate with Meta at all, on old "embrace, extend, extinguish" instincts. Interoperability, it turns out, is a spectrum, not a switch.

Bluesky is the more interesting case, because the company itself is trying to give away its own power. Bluesky PBC has publicly committed to building "a protocol where users can own their data and always have the freedom to leave" — and, more concretely, is partnering with outside initiatives to decentralise the AT Protocol's core data infrastructure onto independent servers, so that neither users nor competing apps depend on Bluesky the company surviving, or staying good. It's the company that built the road trying to hand over the deed before it becomes a toll booth. Nostr goes further still — no accounts on servers at all, just a keypair and a network of interchangeable relays — though it trades away any built-in way to moderate spam or abuse, which is worth saying plainly rather than glossing over.

Who keeps the lights on

Ownership and protocols both answer "who controls this." Neither one answers "who actually maintains it, forever, without either stalling or getting captured by whoever's paying the bills." For that, the clearest working template I've found isn't in transport or social media at all — it's Home Assistant.

The Open Home Foundation is a nonprofit that governs Home Assistant and 250-plus related open-source smart-home projects, funded through a modest paid cloud subscription layered on top of a free, local, fully open core, with three explicit principles: privacy, no vendor lock-in, and staying open to outside contributors. Small foundation, optional convenience layer, open core — it's close to an operating template for what a socialised ride-hailing or social-media protocol could actually look like in practice.

Mastodon is the live experiment in adopting exactly this model, and it's messier and more human than the announcement made it sound. Eugen Rochko said in January 2025 that he'd hand control to a nonprofit, then actually stepped down as CEO that November — citing not crisis but scale and burnout, saying he could no longer see anything about social media without it being about his own work. Felix Hlatky, who'd been consulting on the transition, became Executive Director; the new board includes Twitter co-founder Biz Stone, which is a real attempt at outside accountability rather than a founder choosing his own successor. The entity structure itself is a small, honest mess: a US nonprofit already holds the trademarks, and a planned Belgian nonprofit is meant to replace the original German one — which had quietly lost its nonprofit status back in 2024. Rochko left with a one-time €1 million payment, framed as back-pay for years of below-market salary. Even the good-faith version of "we're becoming a nonprofit" took a false start and a lot of paperwork. Worth remembering before assuming any of this happens cleanly.

The AI argument, and its own catch

Here's the actual, new claim underneath all of this: the reason the first wave of platforms had to be VC-funded, extractive companies is that solving the coordination problem — building and running the software that matches drivers to riders, couriers to orders, hosts to guests, in real time, at city scale — was genuinely hard and expensive. The Drivers Cooperative's own numbers make the case better than any argument I could construct: $600,000 a year in engineering costs is exactly the kind of expense that AI-assisted development is now compressing. Not eliminating — compressing. What used to require a funded engineering team may soon require a much smaller one, which changes who can actually afford to run a cooperative platform without folding after three years.

But I don't want to oversell this, because the same wave that helps also bites. In January 2026, Tailwind Labs — the small team behind the CSS framework that half the web quietly runs on, funded for years by a handful of full-time maintainers and modest sponsorships — cut 75% of its engineering staff after AI coding assistants slashed traffic to its own documentation by 40%, and revenue by 80%. The founder's own warning was that Tailwind could become "unmaintained abandonware when there is no one left employed to work on it." So: AI can lower the cost of building a socialised platform's coordination layer. It can also quietly dissolve the funding model that already keeps some of our shared digital infrastructure alive. Anyone making the optimistic version of this argument owes the pessimistic version equal space.

Gig work should be a choice, not a sentence

One more thing worth saying plainly, because it's easy to leave implicit: none of this requires accepting gig precarity as the price of flexibility. Green Taxi Cooperative is literally unionised. The Drivers Cooperative gives drivers, engineers, and dispatchers all a seat in governance, not just a rating score. Cooperative ownership and full-time, well-paid, professional work are not opposites — gig-app precarity was a design choice a handful of companies made, not an unavoidable feature of matching people who need rides with people who can give them.

This isn't a fairy tale

I'd rather be honest here than triumphant. Stocksy has lasted over a decade; the Drivers Cooperative contracted from thousands of drivers to dozens; an earlier cooperative rideshare attempt in Montreal, Eva Co-op, has an unclear status today. Even Fairbnb.coop's own members, studied in a 2025 paper, weren't sure the platform fully lives up to its own stated fairness. Socialisation is a mechanism, not a guarantee. Plenty of these will fail. The point isn't that cooperative or protocol-based platforms always win — it's that they're now genuinely viable in a way they weren't a decade ago, and that's worth building toward even knowing some of it won't survive.

The road ahead

Go back to the highway image for a second. A road doesn't belong to the company that paved it, and it isn't run by the state either — usually it's maintained by some unglamorous public body, funded by modest tolls or taxes, open to anyone who can drive. Nobody thinks that's a radical arrangement. It's just infrastructure. The argument here is that the coordination layer underneath ride-hailing, delivery, rentals, and social media is infrastructure too, and for fifteen years we let it get built as a toll booth because building a road used to require a company with venture capital behind it.

Varoufakis's fuller proposals — workers holding a real vote over the algorithms running the firms they work for, apps paying people directly for their own data, local media centres run by people picked by lottery — are a much bigger version of this same idea, and I don't think any of them are close to real yet. But CoopCycle is real. Home Assistant is real. The Fediverse, imperfectly, is real. The actual ask isn't to wait for the bigger fix. It's to notice the smaller ones already running, and put some weight behind them.