Farcasterについての考察

For the past few weeks I've been asking myself why I'm still on Farcaster, whether I'll stay, whether I even want to.

I've landed on some answers.

Farcaster, for the uninitiated, was the most credible attempt anyone has made at building a decentralized, crypto-based social network that people actually wanted to use. Founded in 2020 by Dan Romero and Varun Srinivasan, both ex-Coinbase, and backed by $180 million from Andreessen Horowitz, Paradigm, and Union Square Ventures, Farcaster set out to prove that crypto could build something worth using beyond speculation and exit liquidity and the endless recursive loop of tokens that exist to fund the creation of more tokens. It was going to be the social network you actually owned, where your identity and your social graph belonged to you in some meaningful sense, where no single corporate entity could rug-pull your entire online life the way Elon Musk had done to Twitter's culture or Mark Zuckerberg had done to everyone else.

And for a while, it worked. Vitalik Buterin posted there. Developers built interesting things on the protocol; Frames, for instance, let you embed interactive applications directly into posts. The Farcaster team shipped a working decentralized protocol that multiple independent teams could build on without permission. Most crypto projects never come close to that kind of technical achievement. And the vibe was good! If you squinted, you could see the outline of what a post-platform internet might look like: open protocols and communities forming around shared creation rather than algorithmic optimization.

And then 2025 happened.

After the crash

In December 2025, Dan Romero announced that social-first hadn't worked.

Farcaster pivoted to wallets and trading. The thesis was "come for the tool, stay for the network," an honest attempt to find the growth mechanic that the social layer alone hadn't provided. The wallet had been performing well, and Romero called it "the closest we've been to product-market fit in five years." You can argue with the direction, but I don't know that you can argue with a founder who spent half a decade on one approach, acknowleged it wasn't working, and tried something else instead of pumping a token and heading for the exits.

I'd call this integrity by crypto standards and, frankly, by most standards.

In January 2026, Neynar, the infrastructure company that already powered most of Farcaster's ecosystem, acquired the whole thing. Protocol contracts, code repositories, the app, Clanker, all of it. Romero and Srinivasan stepped back // away.

The handoff makes sense. Neynar had been Farcaster's backbone since 2021, serving over a thousand customers. If anyone understood the ecosystem's plumbing, they did. But it also meant the social network now had a new steward, and regardless of how well-intentioned that steward might be, an era had come to an end and the structural reality had shifted.

All of this happened against the backdrop of crypto's broader 2025 reckoning. The memecoin market cap collapsed from $150.6 billion in December 2024 to $39.4 billion by November 2025. The TRUMP token, launched three days before the inauguration with the subtlety of a carnival barker, cratered over 90% from its $75 peak. The LIBRA token, shilled by Argentine President Javier Milei on Valentine's Day, vaporized $4.5 billion and took 86% of its investors' money with it. Over 11.5 million crypto tokens died in 2025, most of them memecoins, most of them launched with no roadmap and no team, with no purpose beyond being the next thing someone could pump before dumping. The October crash wiped $19 billion in leveraged positions in a single event. The Fear & Greed Index, which had read "extreme greed" in September, plummeted to levels that suggested the market had collectively remembered that gravity exists...

If you were looking for a narrative about crypto fulfilling its original promise of financial sovereignty and a more equitable internet, 2025 was a punishing year.

I believe Romero and Srinivasan gave Farcaster everything they had to give. I believe they cared // gave a shit // tried. They spent five years building real infrastructure and shipping real products. They cultivated a community. They weren't exit-scamming or pumping a token. They were doing the boring // unglamorous work of trying to make decentralized social media function at scale, and they ran into the hard problem that it might not be possible.

Loyalty on a sinking ship

Platform loyalty during a platform's decline starts to feel religious. You're maintaining faith in something when the material conditions no longer support that faith. You're posting into a feed that's getting quieter. You're engaging with a community that's growing smaller. The people who remain on a platform in decline (let's be blunt about this) are, by definition, the people who didn't leave, and that group is filtered for stubbornness, ideological commitment, sunk-cost fallacy, or some combinaton of all three.

I know this pattern. We've all lived through it. LiveJournal. Google+. Vine. Tumblr's long twilight after the porn ban. Each one had its diaspora moment, the point where the population crossed some invisible threshold and the network effects reversed. Instead of each new user making the platform more valuable, each departing user made it less valuable, and the departure curve steepened. Robert Metcalfe's law, which tells us a network's value scales with the square of its users, works in both directions. The math is merciless on the way down.

You can map this against Albert Hirschman's framework from Exit, Voice, and Loyalty. When an organization declines, members can exit (leave), exercise voice (complain and try to fix things), or remain loyal (stay and hope). The internet has made exit nearly frictionless (you can sign up for Bluesky in ninety seconds) and voice nearly useless, because platforms at scale have no meaningful feedback mechanism between users and decision-makers. What's left is loyalty, and loyalty without either exit costs or voice mechanisms is inertia.

In Italo Calvino's Invisible Cities, Marco Polo describes a city called Fedora. In the city's museum, there are glass globes containing miniature models of the city, each one representing a version of Fedora that was imagined but never built, all the possible Fedoras that could have existed but didn't. The citizens spend their time gazing at these alternatives, at the roads not taken, the architectures never constructed. Farcaster sometimes feels like one of Calvino's globes: a beautiful model of what decentralized social could have been, preserved in amber, admired by a shrinking group of people who remember what it was supposed to become.

Why I haven't left

I still haven't left.

I wish I had a clean, satisfying reason; something about decentralization principles or the irreducible value of owning your own social graph.

And those things do matter. But the real // honest answer is messier.

Partly it's that the alternatives are all terrible in their own specific ways. X under Musk has become, as Vitalik Buterin put it, "a death star laser for coordinated hate sessions." Bluesky absorbed the X refugees and immediately began replicating the dynamics that made X miserable. Threads is Instagram's vestigial social limb. Mastodon remains Mastodon, which is to say: technically impressive and culturally impenetrable, governed by norms that make posting feel like filing a planning application with the local council.

Partly it's that Farcaster, even in its diminished state, retains something I haven't found elsewhere. The community that remains is small, but it's weighted toward people who build things and people who think carefully about what they're building. The feed isn't optimized for engagement, nobody's trying to go viral, and the conversations that happen there have a texture I associate with early internet forums (in a good way.)

And partly it's that leaving would feel like conceding a point I'm not yet ready to concede.

Crypto's broken promise

The irony of crypto's 2025 collapse is that the technology worked. The Ethereum network processes transactions reliably. Layer 2 solutions have made fees manageable. Smart contracts execute as written. The decentralized exchange infrastructure handles billions in volume. The pipes do what pipes are supposed to do. What failed was the civilization we were supposed to build on top of them. What failed was...well, us.

The crypto pitch I actually gave a shit about (predating the NFT boom and the memecoin casino and the $75 presidential tokens) was infrastructure - building systems that couldn't be captured by any single actor, where the rules were encoded in mathematics rather than terms of service, and where your relationship to a platform couldn't reasonably be compared to serfdom.

That pitch had its origins in the cypherpunks of the 1990s, from Timothy May's "Crypto Anarchist Manifesto" and Eric Hughes's "A Cypherpunk's Manifesto," documents that imagined cryptography as a tool for individual sovereignty in an age of institutional surveillance. The cypherpunks weren't utopians, exactly. They were pragmatists who understood that privacy and autonomy wouldn't be granted by institutions, that they'd have to be built, technically, from the ground up.

Incentives pulled that vision apart. The same cryptographic tools that could enable sovereign identity and censorship-resistant communication enabled speculation at speed and scale. And speculation is both more "fun" than infrastructure and a good deal more viral. It certainly generates better fees, which is why pump.fun, the platform that enabled the creation of thousands of doomed memecoins, remained one of crypto's most profitable companies throughout 2025 even as the tokens it birthed collectivley lost billions in value.

When a technology designed to resist capture becomes the basis for financial instruments, the financial instruments capture the technology. The tail wags the dog. The protocol exists to serve the token, not the other way around. And the people building on the protocol start optimizing for token price rather than for the thing the protocol was supposed to enable.

Farcaster wasn't immune to this gravitational pull. The acquisition of Clanker, the AI token launchpad, in October 2025 signaled a shift in orientation. By the time Romero announced the wallet pivot in December, the trajectory was clear: the social network would become the social layer of a financial product, which is a very different thing than being a social network that happens to use crypto rails. You can respect the pragmatism of that decision (Romero and his team were responding to real data about what users actually wanted) and still mourn the original vision.

A working hypothesis

I cared about Farcaster for the community that decentralization attracted. But decentralization is a means, and means are only as good as the ends they serve. The protocol is the plumbing, and plumbing matters, but nobody moves into a house because the pipes are well-laid.

What Farcaster offered, at its best, was a social environment shaped by a belief in decentralization, a community that self-selected for people who cared about how their tools were built and who controlled them. The protocol was an attractor for a certain kind of person, and that kind of person created a certain kind of conversation, and that conversation was the actual product, regardless of what the cap table said.

This is the distinction: decentralization as architecture and decentralization as culture. The architecture has shifted; the founders have moved on; the wallet pivot reoriented everything toward trading.

But the protocol remains open, and the Neynar team has been embedded in the Farcaster ecosystem from the beginning. They understand what they've inherited. Whether they'll preserve it is an open question, but it's not a foregone co