From a curiosity-driven project to a privacy-first tool built with AI.
Visual Inbox started from a simple question: who am I actually communicating with? Not the words in each message, but the shape of the network being woven, one email at a time. Who stays in the center? Who drifts to the edges? Which relationships are growing, and which are quietly fading?
This version of the idea was built from scratch in 2025, but the curiosity behind it goes back more than a decade. Here's the story.
In 2013, I had a curiosity I couldn't let go of: what would it look like if you could see the social network hidden inside your email? Not the messages themselves, but the structure — who you talked to, how often, and when. I asked two Master's students I was advising at the Media Lab, Daniel Smilkov and Deepak Jagdish, to help me build it. The result was Immersion: a web tool that mapped the relationships buried in your Gmail metadata using only the From, To, Cc, and timestamp fields. No message bodies. Just structure. Daniel wrote his thesis on Understanding Email Communication Patterns and Deepak wrote his on the Immersion platform itself.
The original Immersion demo video (2013)
Immersion launched on June 30, 2013. By coincidence, this was less than a month after the first news stories about NSA surveillance programs broke, and the word "metadata" had suddenly entered everyday conversation. Most people had never thought about what metadata was, and Immersion gave them a way to experience it firsthand — not as an abstract concept, but as a vivid, personal map of their own social life.
The experience was surprisingly emotional. People saw relationships they'd forgotten about, life transitions marked by shifting clusters of contacts, and the quiet drift of friendships they hadn't realized were fading. As one Network World reviewer put it after analyzing his 13.4 years of Gmail and 216,507 emails: the tool revealed the shape of a life. Some users reached out to old contacts they hadn't written to in years after seeing them at the edge of their network.
The project was framed by its creators as being about "self-reflection, art, privacy, and strategy." As co-creator Daniel Smilkov told ABC News: "People are constantly emailing and messaging each other. We wanted to show people that email is a rich and underexplored data field." The TED Blog ran a behind-the-scenes feature, and WBUR (Boston's NPR station) described it as "a self-portrait made from metadata."
"For the press, Immersion was the tool that allowed you to PRISM yourself."
— César Hidalgo, in the Christian Science Monitor (July 2013)
The site crashed within days under the traffic. After relaunching on July 4th, it received over 60,000 visits in a single day. Hundreds of thousands of people used it. The Boston Globe wrote about what metadata says about you, Popular Science covered how it mapped relationships, and NBC News reported on the 70,000+ visitors in its first week. Here are some of the outlets that covered it:
Here are a couple of videos from other creators who explored Immersion at the time:
Immersion is now defunct as a hosted service, but the code remains available as open source. Anyone is welcome to download, fork, and build on it.
View Immersion on GitHubAfter Immersion, I assembled a new team to build something more ambitious: Open Teams. Where Immersion was a personal mirror, Open Teams was a team-level instrument. Multiple members of a team could log in simultaneously, and the system would combine their individual communication networks into a shared view of the team's structure.
The project was built by three people who each brought a different lens. Jingxian Zhang, a Master's student at the MIT Media Lab, did all of the engineering — her thesis, MITeams: Quick Organizational Mapping by Combining Email and Survey Data, documents the technical architecture. Xiajiao Chen, a graphic and UI/UX designer, designed the interface. And Diana Orghian, a psychologist and postdoctoral researcher in my group, designed the questionnaires and psychometric assessments. The three worked together through several iterations to shape the tool.
Open Teams went beyond network visualization. It incorporated several psychological assessments — including personality inventories and collaboration style surveys — that allowed teams to overlay psychological profiles onto their interaction patterns. You could see not just who was talking to whom, but which personality types were bridging groups, where communication silos were forming, and how information flowed (or didn't flow) across the team. It was a tool for collective self-awareness.
Open Teams: visualizing team dynamics and personality types
Open Teams was a considerably larger engineering effort than Immersion. It required server-side coordination for multi-user sessions, a survey infrastructure, richer data models, and a more complex frontend. The project ran for a few years before going offline in 2019. The code remains available as open source.
View Open Teams on GitHubAfter Open Teams went offline in 2019, I missed it. For years, I wanted that view back — the ability to step back and see the shape of my own communication, the network I was weaving through my inbox and calendar. But the old projects were heavy, hard to maintain, and tied to server infrastructure that no longer existed.
So in late 2025, I built Visual Inbox from scratch. This time, I worked with AI as my coding partner, iterating from a blank file to a fully functional suite of visualization tools. The process was highly iterative — start with OAuth, add a network graph, refine the metrics, polish the UI — with each cycle compressing what used to take weeks into minutes. The architecture is fundamentally different from its predecessors:
Immersion and Open Teams processed your data on our servers. Visual Inbox processes everything locally in your browser. Your email metadata, calendar events, and Slack exports never leave your device. There is no backend database, no server-side storage, and no way for us to access your data. This is privacy-preserving by design, not by policy.
To formalize that commitment, Visual Inbox went through the Cloud Application Security Assessment (CASA) Tier 2 verification process. CASA is an independent security review program managed by the App Defense Alliance, a collaboration founded by Google to raise the security bar for applications that access user data. Tier 2 certification means:
This is a level of scrutiny that most personal tools never undergo. It matters because the data Visual Inbox works with — your communication patterns, your relationships, your calendar — is deeply personal. You deserve to know that it's being handled carefully.
Across all three versions of this idea — Immersion, Open Teams, and now Visual Inbox — the core motivation has stayed the same. It's not about surveillance, productivity hacking, or analytics dashboards. It's about seeing yourself.
Your inbox is one of the most complete records of your social life. It captures who you invest your attention in, how your relationships evolve, and what the shape of your world actually looks like — not the shape you imagine, but the one that emerges from thousands of small decisions about who to write to and who to leave for later.
Visual Inbox is a tool for that kind of reflection. I hope it helps you see something about your own life that surprises you, or reminds you of someone you've been meaning to write to.