Thoughts on the Social Graph
I've been thinking a lot about the social graph for awhile now: aggregating the graph, decentralization, social network portability, etc. — If you've seen me at any conference recently, I probably talked your ear off about it.
Context & Ripple Effects
Brad Fitzpatrick turns his conference-circuit argument into a written position: the social graph — who knows whom, scattered across every network separately — should be aggregated, decentralized, and portable between services. The significance is framing: he names a single object ('the social graph') that platforms had each treated as their own private friends-list feature.
The idea did not sit still as theory. Dare Obasanjo's pickup the following day ran under a headline about Google working on a social network aggregator, reading Fitzpatrick's essay less as philosophy than as a description of something a major platform was already trying to build — which is why this post matters beyond its own site.
First-order effects
- The post gives developers and platforms a shared name for the fragmentation problem — one graph, many walled copies — turning 'friends list portability' from a gripe into an agenda with concrete parts: aggregation, decentralization, portability.
- Obasanjo's same-day framing attaches the agenda to Google's reported aggregator work, meaning whichever company builds the index of who-knows-whom first holds the reference copy of relationships rather than any individual network.
Second-order effects
- If a neutral service indexes friendships across sites, networks that treat relationship data as proprietary lose the lock-in that comes from users' contacts living inside their walls — the competitive moat shifts from holding the graph to building better features on top of it.
- Application developers get a new target: build once against a portable identity and graph layer, rather than re-integrating friend-finding and invites for every social network separately.
Third-order effects
- The pattern points toward the social graph becoming shared infrastructure — a layer nobody owns outright, closer to identity protocols than to any one product — though whether it settles into an open standard, a de facto corporate index, or fragments again is genuinely unresolved.
- Networks that refuse interoperability risk sliding down the stack into commodity feature sets running atop someone else's graph, which is precisely the incentive fight that will decide how open the layer becomes.
The trend: Social networking is beginning its shift from closed per-site graphs toward a portable, aggregable identity layer, with large platforms racing to be the one that holds the index.