18 Things Farcaster Got Right. 7 That it Didn't.

Farcaster got a lot of things right—more than almost any crypto consumer project before it. There's a lot of lessons for founders in the space. And the things that went wrong? They're not the things people are complaining about.

Crypto Twitter is doing what it does best: retroactive certainty while also spewing LLM generated rage bait. Everyone knew Farcaster wouldn't work all along, doh!

Most of the takes I'm seeing miss the point entirely. Farcaster got a lot of things right—more than almost any crypto consumer project before it, and there's a lot of lessons for founders in the space. Discussions about money raised or the founders miss the point entirely. And the things that went wrong? They're not the things people are complaining about.

A quick note on me: I built the first third-party client on Farcaster in late 2022 and wrote a number of Farcaster improvement proposals, including ones that led to Channels and Frames. I've been building on or adjacent to Farcaster since then (modprotocol.org, farcaster.id, framesjs.org, openframes.xyz, ethcomments.xyz).

What did Farcaster do well?

1. They started the network with high-quality builders, curating the community largely on the back of Dan's personal network.

2. Dan engaged and dogfooded his product, being permanently online, smart, and a good person to know, listening to all the gripes and ideas (good and bad) from users.

3. They focused on user experience over everything and delivered on it - meaningfully improving over the crypto status quo.

4. They built a world-class engineering team and shipped consistently and fast.

5. They prioritized focus and community-driven participation: they deliberately avoided initiatives like merch, meetups, funding Farcaster public goods, or hosting conferences (until much later). That choice left space for the community to take the lead voluntarily, which proved to be a smart dynamic. They also steered clear of crypto's "free stuff" trap—no grants, no giveaways, no comped conference travel, and similar perks.

6. Dan had a clear view of what was happening across tech and crypto, in part through extensive angel investing that let him see firsthand what was working and what wasn't.

7. They raised enough to be long-term players, and for serious ecosystem players to take them seriously—without this it's much harder in crypto. They nailed the timing of their raises.

8. They unapologetically copied things working in other ecosystems (like Bluesky starter packs)—which is perhaps necessary to win in consumer.

9. They didn't moderate at the protocol layer, avoiding the politics of that problem entirely.

10. They got others to integrate them—particularly as an identity/login method (see Privy)—without getting grouped in with the other 1,000 wallets in WalletConnect, by building a proprietary login method.

11. They didn't allow third-party complexity to creep in, particularly in ways that would affect UX.

12. They pushed a culture of founder-accounts, not company-accounts.

13. They built mobile-first.

14. They leaned into what could make decentralized social on crypto rails 10x better.

15. They didn't launch a token too early without a token having a clear purpose and a way to handle all the spam and farming that would come with it.

16. They leaned into distribution and miniapps as a way to grow and align third-party builders.

17. They didn't let themselves be deceived by stated preferences over revealed preferences.

18. They didn't make themselves dependent or beholden to other pre product market fit companies, nor other crypto ecosystems that weren't at a meaningful scale.

Where I do think things went wrong

Building something valuable is incredibly hard, more so when you're building a network and even more so when doing so in consumer.

It's easy to armchair quarterback, and many of these problems had nuanced tradeoffs in the moment that look different in hindsight. Particularly in the case of social apps, everyone wants the app and network to be built for their needs, but user wants are often divergent and at odds with each other, or at odds with usage metrics.

1. Farcaster didn't have a sufficiently strong network-effect growth strategy

The Farcaster team's thesis was very much product-led growth. Ship features weekly and eventually users will love the product so much they will tell their friends about it. The team wasn't intentional about whether they wanted to grow through onboarding existing communities, building new ones, or neither (purely algorithmic, like TikTok). Beyond that, there wasn't a clear sense of what user cohorts to onboard beyond crypto users.

The product in social networks like Farcaster is the network; the user interface and features are secondary. A pretty Craigslist alternative without the item you're looking for is worse than Craigslist. You need to get all the liquidity to switch at the same time—and Farcaster never succeeded at this, within crypto, let alone attempting anywhere else. This is the unfortunate reality of cold start problems; it's incredibly hard to provide a competitive product in a marketplace with less content liquidity, but more features. Crypto twitter was almost always more interesting, faster to news, and more compelling than Farcaster in the competition of interesting content consumption.

2. Financialization degraded the user experience

As the network grew from 10,000 to 1 million registered accounts (many of which were bots), content quality degraded dramatically, in part due to tokens that rewarded engagement from users in the network, which were heavily farmed. The Farcaster product before 10,000 users was appealing to many because of the quality of builders and founders that new users could engage with—and get replies from. This is one of the hard things about social: some things scale poorly. Spam-free and intelligent replies are a public good, and when there are financial rewards, spam and low-quality content flood the feed and notifications.

The same builders who intelligently engaged in discourse were often the same ones posting low quality replies for the sake of financial rewards. One challenge is how this sort of financial farming can be prevented in a permissionless protocol—even if you hide it from your app, third parties will be incentivised to build toward financial products as a means to monetize.

X recently limited access to their APIs for InfoFi products. A decentralized protocol wouldn't have this lever to deal with this form of financially incentivised content that may be at odds with growing the content quality on it's apps. With LLMs, this issue will only grow as slop becomes harder to identify and easier to create. Token-oriented content is also commonly repulsive to the average non-crypto user, and it makes expansion beyond crypto even harder for teams.

3. Protocol ecosystem neglect

Farcaster kept tight control over the protocol and made it harder for other apps and ecosystems to adopt without being very beholden to the Farcaster team. It's possible that even if they built in a more open way, that no meaningful growth would have come from those apps anyway.

The concept of the "sufficiently decentralized protocol" confused third-party developers like myself, who expected a more conventional crypto protocol but found blurred lines between the protocol and first-party client, which prioritized the Farcaster team's iteration speed over interoperability and consensus. The justification for this was repeatedly that the Merkle Manufactory (the company behind Farcaster) team spending time on allowing others build on top of new primitives would take their time and resources and that it wasn't worth it for them. The only people that could and would be enabled to meaningfully contribute to the protocol's growth was Merkle Manufactory.

The team sourced ideas from third party developers on Farcaster and then competed with those developers via their app. While this may just be the reality of competing in the consumer app market, the power dynamics of protocol & client teams being bundled were such that it either discouraged or outright shut down developers trying to build products and companies on Farcaster.

Another example of this was Farcaster announcing that their app would be open-sourced, and a year later it is still closed source. At some point, the Farcaster team cancelled third-party dev calls. They shipped first-party features in their app repeatedly, calling them protocol changes when they weren't, and also made it impossible for third parties to integrate these features that 90%+ of the network would use (messaging, channels, etc.), making it repeatedly hard for others to build compelling alternative clients.

4. Network topology choices

Some users were crowned with hundreds of thousands of followers by being on the autofollow list for new users, without having to sweat to earn it, which set the network back by shifting the product rapidly away from a game new users could win at, and away from a community-building product—toward a follow-the-existing-influencers product, which I think shifted the incentives for new users to participate.

5. Monoculture

Farcaster couldn't figure out a way to grow distinct communities and content effectively. There were a couple of interations of channels and moderation, but they felt like an afterthought, and experimentation was mostly only possible for the Farcaster team themselves, which I think led to an under exploration of communities as a growth strategy.

6. Capturing downstream value created on Farcaster

Degen and Clanker, both launched on Farcaster and made significantly more revenue than Farcaster, both with very simple products that leveraged the social network as a means to grow.

7. Ownership

The best contributors - from content to builders didn't have a clear path to own any of the value of the Farcaster network they were contributing to. I think this was a missed opportunity to give builders and creators skin in the game.

Where Is Crypto Social At?

Just because activity on Farcaster is down from its peak doesn't mean Farcaster is dead. If anything, the infrastructure and UX standards have been raised for the entire ecosystem and are miles beyond where they were even 3 years ago. LLMs mean 10-100x more apps will be built in the next 5 years. Rish and Manan will take Farcaster in a new, promising direction—a developer-focused direction that is well positioned to leverage the wave of new apps that will be built.

Unsolved Problems in Decentralized Social

• Go-to-market on network effects to overcome the cold start problem. Going from 1 community to many (Coordination problem). Using ownership as a lever for this.
• A way in which the product experience can be 10x better than the status quo, without relying on virtues.
• Monetization of non-financial protocols—particularly to fund public goods and defend against vampire attacks.
• Vampire attack defense. Since pretty much all decentralized social protocol state is forkable, any successful protocol will likely see financial vampire attacks.
• How to measure "quality daily active users" reliably without inviting sybils into metrics—particularly as it relates to getting to network-effect liquidity, testing experiments and rewarding contribution.
• Durable alignment of third-party apps to build and continue to build on a social protocol, even when owning significant market share; There is an incentive for apps to not share all data, or share fake signals to have an algorithmic edge.
• Low-cost data availability, at scale, with low latency, without introducing upfront financial friction for users, and without harming UX or availability guarantees.
• Censorship resistance (at the account and distribution layer).
• Any app-layer innovation will just be copied by apps with more distribution—disincentivizing builders from building fringe apps that try to innovate—and centralizing to the few who can actually try things and benefit from them.
• How to make ownership a lever in growth without all the negatives (attracting the wrong sorts of people, attracting spam, farmers, token talk, distracting the team, having to appease holders in the short term).
• User behavior habits in discretionary-time-spending apps trend toward high frequency in one app (open app to get dopamine)—very centralizing. Apps and distribution are typically owned and controlled by a company, not a protocol. (How to protocolize frontends)
• "Doing things the right way" as a protocol also invites a lot of bureaucracy and overhead for consensus, which is not how innovation happens—particularly in low-financial-risk, highly uncertain consumer products.
• Syncing unread notifications across clients.
• Syncing mutes (and other private, potentially sensitive signals) across clients.
• Spam and low-quality content filtering is still hard—and getting harder with LLMs producing content at scale.
• Interoperability limits experimentation with new things that won't work natively in other apps or will be broken.
• Client-agnostic links (whereby the user sets their client of choice and links auto-open in it).
• How to avoid permissionless crypto rails leading to a casino-like builder and user monoculture that doesn't scale (you can't stop people from building token launchers and infofi—X can block infofi products from its APIs, decentralized social can't).
• Building "the right way" with decentralized building blocks is still incredibly time-consuming, hard, and requires UX tradeoffs. For example, ENS subdomain registries aren't easily indexable right now, tagging users is fragmented, subdomain based namespaces are more complex for users to understand, and more.
• How to have many great decentralized building blocks get built when value capture isn't evenly distributed across primitives (it's easier to monetize off of premium usernames than follows, for example).
• How to help new users onboard and get to the key activation moment faster.
• Enabling experimentation while having interoperability without deep investment and breaking APIs.

Some Opportunities in Decentralized Social

A reminder that distribution rules everything around us, and that overcoming the cold start problem remains the hardest problem for any social network.

Many of the opportunities below may have a product wedge, but will still require tremendous execution and effort to overcome cold start challenges.

• Personalized interfaces and next-gen interaction (actions, apps in feed).
• Next-gen personalized algorithms. Not just ML, but LLMs checking for fit with user interests.
• New platforms and formats (AR, AI).
• Apps as content, private + secure capabilities as differentiation.
• AI-based interfaces for finding and engaging with interesting content.
• Private discounts via AI negotiation, personalization and intents.
• Monetization via micropayments instead of ads.
• Email as a starting point.
• Rewarding network growth and user data targeting capabilities via sybil-resilient metrics (see Worldcoin).
• Ultra-focused experiences for single user persona groups, without mixing them. AI won't let you post/show content that doesn't conform to strict criteria—removes slop and narrows the focus of content.
• Trusted permissions sharing—apps/platforms you will actually grant your location to, for example, that will unlock new primitives.
• Expanding to new signals for ML algorithms to train on.
• Open standards and ERCs that make building easier. For example: how do you fund a new wallet for a user and populate their user information when WalletConnect has unreliable UX and Privy is trying to aggregate this in a proprietary way? Shared miniapp standards that pool social content liquidity across chains and apps without being opinionated.
• Speeding up network-effect formation and tighter launch network selection in order to launch new networks with liquidity faster.
• Letting users and builders own part of the network they contribute to.
• Figuring out a sybil-resistant way to measure quality daily active users (so it can be accurately incentivized, with limited value extraction).
• New social protocol business models that take a cut of valuable activity on the network (Token launchers, Community tokens, ...).
• Credibly + provably neutral recommender algorithms (within high trust domains)
• Intent based social.
• Financial social networks (see Robinhood, Interface, etc.)
• Capital formation around communities.
• Curation markets.

Closing Thoughts

Farcaster got more things right than any crypto social project before it. The team was world-class. The product was genuinely good. And still, it wasn't enough.

That's the brutal reality of building networks: you can do almost everything right and still not win. The timing, the market, the network effects—sometimes they just don't converge.

To the builders still in the arena: the problems are hard, but they're worth solving. Decentralized social isn't dead, it's just a hard problem that we haven't solved yet, and AI is accelerating our ability to experiment and try new solutions.