In the early days of social media, a brilliant new feature could make or break a platform. Remember when Twitter's 140-character limit felt revolutionary, or when Snapchat's disappearing messages created an entirely new category of communication? Those days are over. Today's social media landscape operates under a harsh new reality: features are commodities, and network effects are everything.
The rapid commoditization of social media features has fundamentally changed the competitive landscape. When TikTok introduced its signature short-form video format with sophisticated AI-driven recommendations, it seemed like they had built an unassailable moat. Yet Instagram's Reels not only copied the core functionality but leveraged Instagram's existing massive user base to compete effectively. The feature that once defined TikTok became just another tool in the Meta ecosystem.
This pattern repeats across the industry. Stories, live streaming, voice messages, AI filters, shopping integrations—every meaningful innovation gets reverse-engineered and deployed by competitors within months. The technical barriers to copying features have collapsed thanks to standardized development frameworks, abundant engineering talent, and the open nature of user interface patterns.
What can't be easily replicated is the social graph—the intricate web of relationships, connections, and communities that users have built over years. This represents the true competitive moat in modern social media. When users consider switching platforms, they're not just evaluating features; they're contemplating abandoning their entire digital social infrastructure.
The cost of rebuilding social connections from scratch is enormous, both in time and effort. Users must convince friends to join, recreate their follower networks, and rebuild the algorithmic understanding of their preferences. Most people simply won't make this investment unless the new platform offers revolutionary, not incremental, improvements.
Airchat's brief moment in the spotlight perfectly illustrates this dynamic. The voice-based social platform offered a genuinely novel approach to social interaction, combining the intimacy of voice with the convenience of asynchronous communication. Early adopters praised its authenticity and the unique connections it fostered.
Yet despite positive initial reception and backing from notable investors, Airchat failed to achieve sustainable growth. The platform couldn't overcome the fundamental challenge: convincing users to rebuild their social networks for a feature that, while interesting, wasn't compelling enough to justify the switching costs. Users enjoyed the experience but weren't willing to sacrifice their established Instagram, Twitter, or TikTok communities for it.
Instagram's success with Reels demonstrates how existing social graphs can neutralize innovative competitors. When Instagram launched Reels, they didn't need to build a new user base from zero. They already had over a billion users with established friend networks, follower relationships, and content consumption patterns. Reels simply gave these users a new way to interact with their existing communities.
This network effect created an immediate competitive advantage that TikTok struggled to counter. Users could share Reels with friends who were already on Instagram, creators could leverage their existing follower bases, and the algorithm could build on years of user behavior data. TikTok's head start in short-form video meant little when Instagram could offer the same functionality to users without requiring them to abandon their social infrastructure.
Today's social media landscape operates under different rules than the platforms that defined Web 2.0:
Scale Beats Innovation: A mediocre feature on a platform with a billion users will typically outperform a superior feature on a platform with a million users. The network effect multiplies the value of even incremental improvements.
Distribution Trumps Product: The best product doesn't win—the product with the best distribution does. Existing platforms can push new features to massive user bases instantly, while startups must slowly build awareness and adoption.
Switching Costs Are Social, Not Technical: Users don't leave platforms because they're technically inferior; they leave when their social circles migrate elsewhere. This creates enormous inertia that favors incumbents.
This shift has profound implications for anyone looking to build or invest in social media platforms. Simply having a better feature set or user experience is no longer sufficient. New platforms must either find ways to leverage existing social graphs (like BeReal's integration with phone contacts) or target completely underserved markets where network effects haven't yet solidified.
The most successful recent social platforms have succeeded by either integrating with existing social infrastructure or creating entirely new categories of social interaction. Discord succeeded by focusing on gaming communities that weren't well-served elsewhere. Clubhouse briefly thrived by creating a new format (live audio rooms) during a unique moment (the pandemic) when people were hungry for novel social experiences.
This doesn't mean innovation is dead in social media—it means the bar for success has risen dramatically. New platforms need to offer transformational, not incremental, value to overcome the switching costs associated with rebuilding social graphs. They must solve problems that existing platforms fundamentally cannot address, rather than simply executing existing solutions more elegantly.
The social media giants understand this dynamic and have optimized for it. They maintain massive engineering teams dedicated to rapidly cloning competitor features, ensuring that any successful innovation quickly becomes table stakes across the industry. In this environment, the real competition isn't about who can build the best features—it's about who can maintain the most valuable and defensible network effects.
The era of feature-based social media differentiation is over. What remains is a landscape where network effects reign supreme, switching costs are prohibitive, and only truly revolutionary platforms can challenge the status quo. For entrepreneurs and investors, this means the playbook for social media success has fundamentally changed, and yesterday's strategies are today's recipes for failure.
Share Dialog
Daniel Fernandes
<100 subscribers
Support dialog
Occasionally, I like writing long casts in my own voice, but sometimes I just braindump into a Claude chat window and let it do the expounding for me, this is one of those:
Summary below: https://farcaster.xyz/paragraph/0x297dde1c
Thoughts on zkTLS to force social graph exportability onto web2 giants?
Yes. A couple thoughts: zkTLS could be one tool in the toolbag. But the point of zkTLS is to make any HTTPS webpage into an onchain oracle. In many cases the problem is UX friction that makes the experience suck: OAuth flows, API rate limiting, Keybase-esque "verifying my account" tweets, etc. Anything above "upload my contact book" is a heavy lift. The problem isn't technology. For a highly influential Twitter influencer, every second that they spend exploring a new social media app is a second they could have spent on Twitter, so it's -EV for them to try anything new. Social graphs are idiosyncratic. Some KOLs just got lucky by being early and they would never become famous starting from 0 again, these ppl will never switch. Some groups are on Twitter out of convience and have no loyalty, (thinking of scientists, politicos, etc) and were the first to jump & try Mastodon/Bluesky. Same goes for Reddit. All in all, I think we just have to do the hard thing of slowly building a better social graph...and this would be better done if farcaster and bluesky were working together instead of apart.
I'd counter that zkTLS-proved social connections don't have to be proven by either end of any given social graph edge; any account (or one-off scraper script) could prove the existence of an edge, even between accounts for which onchain entries don't exist yet. This way, it could be feasible to recreate a graph with phantom nodes but non-phantom edges. Then the Keybase-esque flow would only be needed to claim a phantom node. I feel like that would meaningfully reduce the onboarding friction of individual users.