Common Notes is implemented through AIs writing most notes, users writing and voting on notes, and a browser extension that shows the system on any site, including ACX.
Unresolvable The public record could not settle it.
Factual: Statements about the present or the past, checked against the record: did they hold up?
What would count as true or false
Judge as of 2026-10-19 against Common Notes' own documentation, launch materials or independent coverage. The claim has three parts: (1) AIs write most notes; (2) users write other notes and vote on which are helpful; (3) a browser extension makes notes appear on arbitrary websites, including Astral Codex Ten. True: all three are confirmed, including that AI-written notes are the majority or are designed to be. Mostly true: all three mechanisms exist, with a minor slip in a detail, e.g. AIs write many but not clearly most notes, or users rate notes rather than strictly 'vote'. Mixed: one of the three parts is missing or materially different, e.g. there is no browser extension, or AIs write only a small share. Mostly false: the project exists but two of the three mechanisms are absent or materially different. False: no project called Common Notes worked this way at that date, e.g. it is purely human-written with no extension.
Set before the research began.
Evidence (8)
-
E1 R Street Policy Study: AI Writers on Community Notes: An Evaluation of Seven Months of Data R Street Institute · 16 June 2026
“Over the full study period, AI’s share of total monthly note volume rose from 0.2 percent (September 2025) to 12.7 percent (March 2026).”
-
E2 R Street Policy Study: AI Writers on Community Notes: An Evaluation of Seven Months of Data R Street Institute · 16 June 2026
“Human note volume remained relatively stable, ranging from approximately 54,000 to 70,000 notes per month.”
-
E3 R Street Policy Study: AI Writers on Community Notes: An Evaluation of Seven Months of Data R Street Institute · 16 June 2026
“Rather than relying on a simple majority vote, which could allow partisan blocs to promote favorable notes or suppress unfavorable ones, the algorithm models each rater’s ideological position”
-
E4 R Street Policy Study: AI Writers on Community Notes: An Evaluation of Seven Months of Data R Street Institute · 16 June 2026
“CRH is the key outcome in the Community Notes system, as only CRH notes are visible to all users on the platform.”
-
E5 AI Writers on Community Notes: An Evaluation of Seven Months of Data R Street Institute (R Street Policy Study No. 338) · June 2026
“combined 31,464 notes, or 7.4 percent of total volume of all notes.”
-
E6 AI Writers on Community Notes: An Evaluation of Seven Months of Data R Street Institute (R Street Policy Study No. 338) · June 2026
“and a separate pool of contributors rates those annotations for helpfulness.”
-
E7 AI Writers on Community Notes: An Evaluation of Seven Months of Data R Street Institute (R Street Policy Study No. 338) · June 2026
“This system, designed entirely around human participation and judgment, was the status quo until late 2025.”
-
E8 Links For October 2026 Astral Codex Ten (Substack) · 29 October 2026 Own record
“Remember to ask an AI about any surprising article you read, or use the browser extension Common Notes so other people can do this for you.”
Reasoning
E1–E7 come from an R Street study of X's Community Notes, a different system. They say nothing about how Common Notes works, so they can't confirm or refute any of the three parts of the claim. The only source that names Common Notes is the subject's own blog post (E8). It calls Common Notes a browser extension but does not show that AIs write most notes or that users vote on which notes are helpful. No Common Notes documentation or independent coverage was provided, so the claim can't be settled.
Cite this claim
Scott Alexander, “Common Notes is implemented through AIs writing most notes, users writing and voting on notes, and a browser extension that shows the system on any site, including ACX.” (said 19 October 2026). Verdict: unresolvable. Reliability API, claim c_543ea25f54ad, method 0.1.0+8bea457b, dataset release v1.1.2. https://epistacklabs.com/claims/c_543ea25f54ad
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More factual claims by Scott Alexander
-
Mostly true
“MNX is a real-money prediction market on AI.”
-
Resolved true
“Friend of the blog Goodheart Labs is working on Common Notes, aka “Twitter Community Notes, but for everything”.”
Goodheart Labs is working on Common Notes, described as Twitter Community Notes but for everything.
-
Mostly true
“Reader Nikita Sokolsky has completed his work getting some extra Fatima documents from Portugal and translating them into English.”
-
Mostly true
“Physicians For AI Safety wishes to advertise that they exist, are interested in talking to any physicians interested in AI safety, and have an open letter that such people can sign.”
-
Resolved true
“Rob Ennals, who made this year’s book review contest site, has gathered the reviews from earlier years into epubs here.”
Verdicts are made automatically; how verdicts are made.