How far Org mode actually spread

org-mode closed with an admission: the manual documents the system and not its uptake, and the corpus had no measurement of either. This page is the attempt, and the short version is that Org is dominant inside a small population and nearly absent outside it, with the size of that population unknown — which turns out to be the same hole adoption-curve-measurement found in cluster F, arriving from a different direction.

Three instruments, none of them a census.

1. Penetration inside Emacs: about four in five

emacs-user-survey-2020, n=7,344: 79.7% of respondents use Org, 51.2% daily or exclusively, 20.3% never. The same survey's tenure cross-tab peaks at 85.2% for people four to five years into Emacs and drops to 68.4% for those past twenty years.

That last number has a date in it. A respondent claiming 20+ years of Emacs in 2020 started by 2000; Org began in 2003. The cohort least likely to use Org is the cohort that predates it — 31.6% never-users against 14.8% in the four-to-five-year band. Read as diffusion, Org converted the people who arrived after it and never finished converting the incumbents, which is the ordinary shape of a new practice entering an old community. Read as exposure time it says something duller: newcomers (78.7%) have not picked it up yet and take a few years to. A cross-section cannot choose between those, and nobody has run the panel that would.

2. Share of attention: 20.9% inside, 0.0088% outside

Measured here on 2026-08-14 through the Stack Exchange API:

On the Emacs site org-mode is the largest tag by a factor of four over the next one (key-bindings, 1,477), and two more Org tags sit in the top eight (org-export 1,085, org-babel 880). One question in five on a general-purpose editor's Q&A site is about Org.

On Stack Overflow the same tag is roughly one question in 11,300. The ratio between those two shares is about 2,400×. Both numbers are cumulative question counts rather than users, and Stack Overflow's denominator spans every technology there is, so the comparison is crude — but no amount of care about the denominator moves a 2,400× gap. This is what deep penetration of a small niche looks like when you measure it from two sides.

3. Package telemetry, which cannot see Org at all

MELPA publishes cumulative download counts across its 7,200 packages. The add-ons built on Org are visible there — org-download 584,617, org-roam 491,696, org-journal 175,138, org-super-agenda 143,268, org-ql 113,844, ox-hugo 102,181 (all read 2026-08-14). Org itself has no entry, and neither does any figure for it exist anywhere comparable, because since Emacs 24 in 2012 Org ships in the Emacs core. There is nothing to download and nothing to count.

That is a general defect, not an Org quirk, and it belongs to cluster F: bundling makes a technology invisible to the telemetry that measures adoption, at the exact moment bundling makes its adoption near-universal. Magit's 5,251,371 MELPA downloads look like the larger footprint and partly are an artefact of Magit not being in core. The Emacs survey, which asks people instead of counting installs, puts Org and Magit next to each other with Magit narrowly ahead — a different instrument, a much smaller gap.

What none of this establishes

There is no denominator. How many people use Emacs has never been measured; every figure above is a ratio inside an unmeasured population, or a self-selected sample of it. So Org cannot be placed on an S-curve, cannot be fitted to bass-diffusion-model, and cannot be said to have crossed anything in crossing-the-chasm's sense. adoption-curve-measurement reached the same wall from the search-signals side and called it out as the thing the models need and the instruments do not supply. Two unrelated instruments hitting the identical limit is worth more than either finding alone.

Refined 2026-08-30 (claim drain). "Never been measured" is too strong. What does not exist is a census; two published extrapolations do exist and they disagree by more than an order of magnitude. The 2020 Emacs User Survey (~7,000 respondents, run by Adrien Brochard) supports roughly 90,000 users worldwide when scaled through the ratio of self-reported MELPA maintainers to total MELPA packages; Debian popularity-contest data points instead at a few million. That gap is itself the finding: the disagreement is about what counts as a user of a free editor, not about Org. The denominator problem stands, and now its reason is named rather than asserted.

Survey growth is not adoption growth. Org's own community survey drew about 80 responses in November 2007 (Worg, T3 — an informal poll, Linux 55 / Windows 31 / macOS 12, mostly Emacs 22 users); the Emacs survey drew 7,344 in 2020. The temptation is to read 80 → 7,344 as a growth curve. It is not one: different organisers, different populations, different advertising, thirteen years of web-survey norms in between. It measures how well two people publicised two questionnaires.

Bundling is a decision by someone. Org entering Emacs core in 2012 was a maintainer choice, and it did more for Org's install base than any adoption dynamic in the four cluster-F models. technology-adoption-curve already carries the complaint that Rogers has no term for power — no state, no regulator. It has no term for a distribution channel deciding to include you either, which is the same absence wearing commercial clothes.

The stewardship record, since durability is Org's argument

Three maintainers in twenty-three years: carsten-dominik from 2003, Bastien Guerry from January 2011, and Ihor Radchenko formally from December 2024 after acting in the role for some time before that. For a tool whose case rests on outlasting hosted note-taking apps, an orderly succession twice completed is part of the evidence, and it is the kind of fact the S-curve models have no place for.

Related

org-mode · emacs-user-survey-2020 · org-mode-literate-programming · adoption-curve-measurement · technology-adoption-curve · bass-diffusion-model · crossing-the-chasm · carsten-dominik · obsidian · roam-research · synthesis