Every chart, computed from data
Nothing here is hand-entered. Every number comes from graph.json, which is rebuilt from the published builders on every build. Small sample: read patterns as anecdotes until the series grows.
3D globe
how this is computed
Pins at each builder's city (rounded to 2 decimals), diamonds at origin. Arcs: origin to home (dashed), person to person where similarity ≥ 0.35 (thickness = score), and Fred to each (toggle on the Atlas).
Similarity heatmap
how this is computed
Pairwise weighted Jaccard over themes (w 2), tools (w 1.5), languages, models, hardware class and previous OS (w 1). Rows ordered by average-linkage hierarchical clustering.
Connection graph
how this is computed
Edges are pairs with similarity ≥ 0.35 plus hand-curated connections (orange). Colors are communities from one-level Louvain modularity: Lee + Michael (cloudflare pages); James + Yasir (old hardware).
Theme × person matrix
how this is computed
Each builder's themes against the synthesized THEMES list, sorted by prevalence. 12 themes were arrived at independently by 2+ people.
Trait correlations
| linux toy until omarchy | ↔ | codex | φ 1 | both: 3/4 |
| looks then ai | ↔ | codex | φ 1 | both: 3/4 |
| ai vampire | ↔ | codex | φ 1 | both: 3/4 |
| memory is the product | ↔ | github | φ 1 | both: 3/4 |
| memory is the product | ↔ | claude | φ 1 | both: 3/4 |
| memory is the product | ↔ | repo | φ 1 | both: 3/4 |
| cost routing | ↔ | codex | φ 1 | both: 3/4 |
| local vs frontier | ↔ | github | φ 1 | both: 3/4 |
| local vs frontier | ↔ | claude | φ 1 | both: 3/4 |
| local vs frontier | ↔ | repo | φ 1 | both: 3/4 |
| github | ↔ | claude | φ 1 | both: 3/4 |
| github | ↔ | repo | φ 1 | both: 3/4 |
how this is computed
Phi coefficient between binary traits across builders. Listed only when n ≥ 4. n=4: anecdotal, read as patterns to watch, not findings.
Stack leaderboards
how this is computed
Counts per tool, model, language, hardware class, previous OS and how they found Omarchy. Full boards on /stack/.
Previous OS → Omarchy
how this is computed
Flows from each previous OS (split evenly per person) to the class of their first Omarchy machine, then to their daily driver.
How long the road was
how this is computed
First computer year, Linux since, Omarchy since: one line per builder.
Series growth
how this is computed
Cumulative published interviews by date, with 18 booked calls dotted ahead (count only, as of 2026-10-13).
Countries and timezones
how this is computed
Home countries: US 2, CA 1, SG 1. Origin countries: GB 2, PK 1, US 1. The series spans 14 hours of UTC offsets.
Quote wall
“I never trust, um, one—one AI. I always get one to check the other.”
Lee Shand · L427
“turn it into issues, and then go do a PR stack, solving each issue in sequence.”
Michael Hurley · L198
“I'm s*** at implementation”
Lee Shand · L463
“I just built s***, right?”
James Tervit · L965
“I just went in always with the attitude that this is a tool that I'm going to use to my advantage.”
Michael Hurley · L37
“I love to design much more than I love programming, but I'm good at the programming thing.”
Michael Hurley · L154
“there's probably, like, almost 200 repos on there now since January.”
Michael Hurley · L169
“I think I stopped learning in 2004”
Yasir Malik · L189
“And I don't like being dependent on someone.”
Yasir Malik · L195
“my Mac can literally just stay closed in the corner.”
Michael Hurley · L301
“This is how it used to be on BBS with community. Everybody was sharing”
James Tervit · L374
“He's revived the whole, um, thing about community”
James Tervit · L1020
“this thing got me hooked on how I used to be when I was a kid”
Yasir Malik · L75
“the way that I've been, um, significantly more productive has been when I installed Omarchy”
Lee Shand · L360
“if I don't know any other application that's better than the terminal, um, it's my go-to.”
James Tervit · L359
“Linux felt like a toy until Omarchy 3, really. And then it was, okay, I can use this for work.”
Michael Hurley · L27
“For people like me, if I don't have AI anymore, I don't have a life.”
Lee Shand · L90
“That's where all my cognitive function lives, is within Obsidian.”
Lee Shand · L82
“My goal with Obsidian is I never want to be in that application”
Lee Shand · L78
“This is— this is just fun”
James Tervit · L350
“I just don't feel there are enough hours in a day. That's my only complaint.”
Yasir Malik · L259
“I'm so glad that I'm living in this day and age where technology is just like moving at the pace it is.”
Yasir Malik · L123
“To be honest, the looks were the first thing that got me.”
Yasir Malik · L68
“With Omarchy, any issues that I had, it's AI is built in.”
Yasir Malik · L84
“I better get my finger out because I saw your website.”
James Tervit · L184
“how you almost don't have to think to use it”
Lee Shand · L154
how this is computed
Each builder's quotes, grouped by the theme they illustrate. Every quote links to its transcript line.
Trait radar
how this is computed
Share of indicators each builder has on six axes (agents, local AI, design, data, hardware, accessibility), against the community mean. Indicators live in graph.config.yml.
Most connected, rarest
how this is computed
Most connected: Michael Hurley (degree 3, strength 1.207). Items used by exactly one builder: blip (Michael), cesium (James), claude-code (Lee), cloudflare-tunnel (James), herdr (Yasir), hermes (Lee), nuphy-node-100 (Yasir), obsidian (Lee), omadesign (Michael), proxmox (James), readwise (Lee), t3code (Michael), vercel (Michael), gpt (Yasir), grok (Michael), bash (James), html-css (Michael), python (Yasir), rust (Michael), sql (Yasir), typescript (Michael), vba (Yasir).
Talk time and distinctive words
how this is computed
Share of words spoken by the builder in their cleaned transcript (they should dominate), and top TF-IDF terms per builder over the cleaned transcripts (stopwords removed, terms said 3+ times).