Nonpartisanship and editorial independence
We publish sourced records of what people in public office did. We do not endorse, oppose, rate, or rank anyone standing for election.
ConnectRealDots is an independent publication, currently run by its founder. A Washington nonprofit corporation and an application for 501(c)(3) status are in progress and neither is complete. This page states the standards we hold ourselves to now; where something depends on a step we have not taken yet, it says so. Contributions are not tax-deductible, because we are not yet a charity and are not asking for any.
What we publish
- What officeholders and institutions did — orders signed, votes cast, laws passed, judges confirmed, money spent, cases filed.
- What they said that can be checked — whether a pledge can be paid for, whether it is legal, whether it has been done before. Never its effect on an election.
- Public data over time, from government and academic sources.
- Every figure traced to the record it came from, so you can check it rather than take our word.
What we do not publish
- Endorsements or opposition, express or implied.
- Candidate ratings, scorecards, rankings, or comparisons of who is better.
- Voting recommendations, or any suggestion of how you should vote.
- Predictions or commentary about who will or should win.
- Anything timed to an election rather than to the event it describes.
We do take positions on public policy — that a figure is rising, that a claim is unsupported by its source, that a mechanism does not explain what it is said to explain. That is the point. What we do not do is turn a policy position into a message about a candidate.
Why you can check this rather than believe it
Any publisher can claim to be even-handed. The reason to believe this one is that the even-handedness is built into the machinery, and the machinery leaves a record.
- The same spec, applied to both parties. Our pages on named politicians cover both, built to one published specification with the same metrics, windows and sources.
- A rule that has caught us favoring people. Every entry must name an action taken by the subject. That rule has caught real errors in all three records we have built — crediting one country's treaty suspension to a president, and a Senate rule change to two others. It finds errors in both directions and has been applied to both parties.
- Figures are corrected upstream. When a number on a person's page disagrees with the topic page it comes from, we fix the source. It is not possible to quietly improve or worsen one subject's numbers.
- We withdraw our own published claims. A second automated reader re-examines what we published and can correct a figure or withdraw an entry without a person asking it to. Those withdrawals are public.
The test we hold ourselves to
Charitable organizations are prohibited from intervening in a political campaign for or against a candidate. The IRS applies a facts-and-circumstances test, set out in Revenue Ruling 2007-41. We use its factors as our own editorial checklist, because a rule you can check beats a promise you cannot.
| The factor | Where we stand |
|---|---|
| Does it identify a candidate? | Often yes, unavoidably — we write about officeholders, and officeholders stand for election. This factor alone is not disqualifying; the other six carry the analysis. |
| Does it approve or disapprove of a candidate? | No. We report what was done and what it measures. Our judgments are about evidence — whether a figure is supported by its source — not about whether an action was good. |
| Is it close in time to an election? | Sometimes, because we publish every day. The next two rows are why that is not the problem it appears to be. |
| Does it refer to voting or an election? | When the subject is election administration — a ballot rule, a court case about one — or a pledge that depends on an election result, which we check for its substance (its cost, its legality, whether it has been done before) and never for its effect on the race. We never invoke an upcoming election as a reason to read something. |
| Is the issue one that divides candidates? | Sometimes. Public policy is contested; that is why it is worth documenting. The ruling is explicit that a charity may address issues that divide candidates. |
| Is it part of an ongoing series, independent of election timing? | Yes, and this is our strongest fact. We publish on a fixed daily schedule that predates this election cycle and does not change for it. |
| Is the timing tied to a non-electoral event? | Yes. Every entry is triggered by a documented official act or public statement — an order signed, a suit filed, a figure released, a pledge made — and is dated to it. |
The last two are load-bearing, and we protect them deliberately: our publishing schedule must never become election-driven. Timing coverage to a campaign would forfeit the two facts that most clearly separate what we do from campaign intervention. That is a standing constraint on how this site is built, not just on how it is written.
Automated publishing
Much of this site is produced by automated systems that select, draft, check and publish without a person approving each item. We say so plainly, because concealing it would be the actual problem.
That automation is governed, not unsupervised. The daily brief on named officeholders makes the most claims about people, so it is governed most tightly:
- A deterministic validator rejects malformed or unsupported claims before any model sees them.
- An editor checks each draft against its sources and drops what it cannot verify rather than rewriting it. If the editor cannot run, nothing publishes.
- A campaign-intervention check applies the factors above and drops any draft that reads as support for or opposition to a candidate. A pledge that depends on an election result is kept only when it is checked for its substance, shows where experts disagree, and says nothing about its effect on the race.
- A second reader re-examines published entries and can correct or withdraw them.
- An anomaly guard stops the machinery and escalates to a person when the checkers themselves start behaving oddly.
Every step of the brief fails toward publishing less, never toward publishing something unchecked.
Daily Dots are published first and checked straight afterward, so they work differently:
- An editor checks figures, dates and names against sources, corrects what is wrong and removes what it cannot verify. A correction is applied only after a second, adversarial reading approves it.
- The same campaign-intervention check reads every Dot. Framing that reads as support for or opposition to a candidate is removed, and where it is not removed the Dot is not posted to social media.
- A Dot is posted to social media only after that check has finished. If it cannot finish, the Dot stays on the site but is not posted, and the failure is recorded. The same is true when the check finds an error and the correction does not fully go through.
Independence
We accept no advertising, take no sponsorship, and sell nothing. No funder, donor or partner reviews or approves coverage before publication, and none may direct it.
We have taken no restricted funding. If we ever do, the restriction will be disclosed on this page. Once the organization is incorporated, its board will govern it and will not edit it — the same separation, written down before there is anyone to enforce it against.
Corrections
We correct in public. A corrected figure shows the old text alongside the new; a withdrawn entry stays visible and says it was withdrawn. We do not quietly delete.
If you believe we have published something inaccurate, or something that reads as favoring or opposing a candidate, tell us and we will look at it. That includes complaints about this page.
Review
We re-read what our automated systems published about anyone on a ballot, and confirm the campaign-intervention check still works in both directions — blocking advocacy, and not blocking a sourced record of what an officeholder did. That review runs at least once a year and in the January of any federal election year. Once a board exists, each review will be recorded in its minutes.
Last reviewed 1 September 2026 · privacy