A digital-archival exploration · Generative Anthropology
Between 2019 and 2022, a community formed around a note-taking app and called itself a cult. It was unmistakably mimetic — yet it never produced a scapegoat. What did it produce instead?

People tattooed a note-taking app's logo onto their bodies. A believer's leg. The founder's own father's arm. By early 2021 the count had reached five — the founder tallying each "#roamcult tattoo" as it appeared. — the most permanent sign a digital scene can leave
Follow the research on Substack ↗
"Roam is truly making the 'second brain' meme not cliché… notes organized in the shape of thoughts… a search bar to your memory. Listen, I'm this close to getting a mariner's astrolabe tattoo. Please, you have to believe me. You can't compare it to Notion."— @abettertake, Jan 2021 · source ↗
Photographs are public posts, reproduced for scholarly commentary with links to source. Faces of non-public individuals may warrant consent or cropping before wider publication.
What began as a user community around Roam Research rapidly took on characteristics its own members recognized as cult-like: fervent testimonials of cognitive transformation, evangelical recruitment, tattoos of the logo, public sharing of personal knowledge graphs, and a tiered [[Believer]] membership that formed an inner circle around the product's development. The intensity peaked as pandemic lockdowns drove social life online. By 2022, the fervor had largely dissipated.
This is a hypothesis-generating account, not a completed ethnography. It draws on public Twitter/X discourse, community traces, and the researcher's own participation during the period of greatest intensity. It asks whether Roamcult can be understood as a case of scenic formation — and whether Eric Gans's originary thinking can clarify what René Girard's scapegoat mechanism alone does not.
Roamcult was clearly mimetic. Participants did not arrive at enthusiasm through isolated evaluation of features; they encountered others' visible desire and were drawn into it. Scrolling Twitter, prospective users saw life-changing testimonials, shared workflows, screenshots of personal graphs. They desired not merely a useful tool but the clarity, belonging, and identity others appeared to have found. Desire spread through contagion.
Yet Roamcult did not culminate in sacrificial crisis. The community generated collective energy without being organized by destructive rivalry. Participants outdid one another in helpfulness — tutorials, public thinking, plugin development, zeal. One participant's phrase captures it:
"friendly ambitious nerds"— a Roamcult participant, on the community's competitive generosity
If Girard explains the contagion of desire, Gans may explain how that desire was organized around a shared center. In the originary hypothesis, the group constitutes itself through shared reference to a center of attention that both intensifies and regulates desire. Here, Roam-as-represented functioned as such a center — not simply software, but a shared object of transformation around which participants organized practices, identities, and mutual recognition. The technical substrate (bidirectional links, block references, daily notes, graph structures) was real; but the center was constituted by collective representation: the claim that Roam changed how one thought.
The tongue-in-cheek self-identification as a "cult" may have intensified this scenic formation. The ironic framing licensed forms of fervor that might otherwise have felt excessive — participants could perform zeal while staying in on the joke. But observers could read those performances as earnest, and public signs of devotion fuelled further imitation. What began as irony may have accumulated into sincerity.
Mimetic desire is drawn into a common field of attention organized around Roam-as-represented.
Participants orient toward the same center with escalating intensity, held in productive tension by shared signs and recognition.
Obsidian and Logseq draw attention elsewhere; pandemic conditions ease; the company's public energy diminishes. The scene dissipates — without a scapegoat.
Roamcult is offered as a pilot case for digital scenic formation. Platforms make others' desires continuously visible, accelerating contagion but also making scenes fragile: attention gathers quickly around a center, and migrates quickly when the center loses charge.
— Is "sacred center" the right Generative Anthropology concept for Roam-as-collectively-represented?
— How might one distinguish mimetic contagion from scenic formation using public digital archives?
— What interviews, artifacts, or comparative cases would turn this pilot into a fuller ethnographic project?
The archive · 2019–2022
A corpus of public Twitter/X posts about Roam Research and #roamcult, collected by keyword and by account, deduplicated into a single archive. Below: the patterns the argument rests on, and the full archive to browse yourself. The traces show how enthusiasm, identity, and shared attention became publicly legible — not the private experience behind them.
The scene, moving
Two ways to move through roughly ten thousand tweets in time. One looks down on the scene from above; the other stands inside the stream of voices. Both show the same arc — a center swelling into coherence, then losing its charge as attention migrates.
Each tweet is a point of light, clustered and coloured by theme, sized by likes. Scrub the timeline and watch the gold cult-and-sacred cluster swell toward early 2021, then watch the violet rival-tools cluster light up as the centre fades.
Best viewed large. Open this view fullscreen ↗ · Themes are assigned by a transparent keyword classifier; drift speed in the immersion view is a deliberate interpretive choice, not a neutral replay.
Read it · annotate it
The visualizations let you feel the scene move. These let you read it slowly and leave notes anchored to specific moments. Each is a Google Doc with commenting open to anyone with the link — select any tweet, hit comment, and your note stays pinned to that spot in the timeline. Together, readers' comments become a shared commentary track.
The spine of the conversation: the highest-signal tweets, drawn from the community's own accounts and its self-naming language, arranged in time. A start-to-finish walk that takes a single sitting — a community gathering around Roam Research, peaking in the pandemic, then dissipating as attention moves on. The best entry point if you're new to the scene.
Open & comment ↗ For immersion · the full archiveThe complete curated archive — every tweet from the accounts at the heart of #roamcult — laid out one day at a time. Built for immersion: read a day or a week at a sitting and let three years pass as weeks, recreating what it was to live inside the scene as it formed and faded. Each day is a heading in the outline, so you can pace yourself and find your place.
Open & comment ↗Both open in Google Docs with comment access for anyone with the link. ◆ marks the center (Roam / the founder) speaking; a gold bar flags high-engagement tweets. Theme tags come from the same keyword classifier as the visualizations — a rough guide, not ground truth, and a fair thing to argue with in the margins. The "view on X" links resolve to the original tweets where still public.
Qualitative readings
Counts show the shape of the scene; specimens show its texture. Below: individual posts read closely against the argument, and moments where one person's visible desire was re-broadcast by others — mimetic circulation caught in the act.
A quote tweet is desire with a signature on it: re-sharing what someone else made visible, plus one's own endorsement. These are reconstructed from quote links preserved in the archive.
Tongue-in-cheek, and not
Roamcult performed its own cult — half as a joke, half in earnest. That doubleness isn't a sideshow to the argument; it is the argument. The community supplied the religious vocabulary itself, then meme-ified its own devotion. Here's the bit as the archive caught it — and the turn that community memory says ended it.
None of this is outside commentary — it's the community doing the bit on itself, in earnest enough to be felt and ironic enough to be deniable. Each card below is the exact text of a public post; the grey italic line beneath is my annotation, not theirs.
By participants' own telling, the late phase pointed the irony back at the center. A half-joking, half-serious move — the "roaminati" — turned on the founder himself, and former Believers drifted off to Obsidian and Logseq. The drift is visible in the data; the story behind it lives in community memory, not the public archive — the kind of thread the interviews exist to follow.
How the archive was made — and the raw data itself
The argument above is interpretive; this page documents the evidence honestly, including where it is incomplete or noisy. Transparency about the archive is part of the method — and the raw data is yours to take.
The complete dataset (12,206 tweets, 2019–2022) in machine-readable form: the deduplicated master file plus the keyword and account subsets, as JSON and CSV. Many of you are better with data than I am — pull it apart, chart it your own way, run your own analysis. If you find something I missed, I'd love to hear about it.
Download the raw data ↗Public posts from X (Twitter) spanning were collected along two paths. A keyword path searched hashtags and Boolean queries (e.g. #roamcult, (roam OR roamcult) (template OR workflow OR graph …), and a query targeting the "Believer" pricing tier). An account path scraped the Roam-related posts of the founder, the official account, the self-styled cult account, and a set of active community members. The two were merged and deduplicated into one master archive of unique tweets.
The merge is clean: every tweet has a unique identifier, with no duplicates and no missing core fields, and dates and links are well-formed. The known weakness is the keyword path. Because "roam" is an ordinary English word and "believer" an ordinary phrase, the broad Boolean queries — the "Believer" query especially — pulled in false positives: VR-gaming "free-roam" promotions, "true believer" political and religious posts, a Metallica lyric, and similar noise. A conservative filter flags the clearest of these as off-topic; they are hidden in the browser by default and excluded from the charts. Borderline "Believer" cases remain and should be hand-checked before any quantitative claim about that term.
The proposal locates peak intensity in "the early months of the COVID-19 pandemic." The #roamcult volume curve instead peaks in December 2020 – February 2021, roughly a year into the pandemic. The honest reading distinguishes onset (early-pandemic conditions seeding formation in 2020) from peak intensity (early 2021) — and the lag is itself interesting: scenic coherence took months of accumulating visible desire to crystallize.
Public archives can establish the shape of the phenomenon — its arc, its center of attention, the migration of interest. They cannot, on their own, adjudicate "mimetic contagion" versus "scenic formation," since both predict similar surface traces; that distinction is theoretical, turning on the kind of organization rather than on counts. Nor can tweets reveal private experience. The archive marks the field of visible desire and its center; whether that center is a Gansian sacred center is the question brought to the conference, not a result claimed from the data.
— Early years (2019–2020) are likely incomplete.
— Deleted and protected tweets are not included (survivorship bias toward what persisted publicly).
— Engagement counts reflect values visible at collection time.
— This is one platform; Roamcult lived across Slack, the forum, Reddit, and YouTube as well.
— The scrape did not capture reply linkage (parent-tweet IDs are absent and conversation_id is a stub), so reply-thread reconstruction is not possible; quote-tweet relationships were preserved and are used instead. A future scrape capturing in_reply_to_status_id would enable thread-level analysis.
The presentation
The GASC slide deck and the recorded presentation. (Embeds are wired up — drop in your URLs and they appear here.)
Get in touch
If you were part of Roamcult, or you're working on something adjacent, I'd like to hear from you. Reach me directly, or follow the research as it develops.
Kahlil Corazo · Department of Anthropology, Ateneo de Davao University.
roamcultresearch.substack.com ↗The form uses Netlify's built-in handling — submissions appear in your Netlify dashboard under Forms, no backend needed. It activates automatically once the site is deployed on Netlify.