No Original Research

Articles must not present original research – unpublished analysis, speculation, or conclusions that no reliable source has stated.^[1] Every interpretive claim should be attributable to a published source.

Reporting on-chain data

Reading and reporting a value from the ledger (a balance, a supply, a proposal outcome) is description, not original research, and is encouraged. Extrapolating from it – predicting price, inferring intent, or drawing unpublished economic conclusions – is original research and belongs in the Blog or Ideas Pipeline, not in an encyclopedic article.

Where the line has fallen

Four cases from this wiki’s own record show what separates a reading from an inference. Each sits on a page a reader can open, and in each the handling now published is the correct one.

Correct data, invented model

An earlier version of this wiki’s account of the May 2026 governance votes reported them as falling short of the 671,470,000 XRD quorum written into each. The balances it read were accurate; the model applied to them was not. Voting weight on those accounts is held mostly as stake units rather than liquid XRD – 785,092,070 of the 910,989,217 XRD behind the Stokenet proposal – so counting liquid balances alone understated the electorate more than sevenfold and inverted the outcome. Reading the balances was description. The total assembled from them was an unpublished model presented as a measurement.

A silence is not a reason

Hermes Protocol is recorded as dormant on four dated checks: the Enter App button on its own homepage points at a host with no DNS record, its announcement channel has been silent since June 2023, its newest commit is from February 2024, and its website still serves normally. The page states those four things and then states the gap – no wind-down was ever announced. Why the project stopped is not established, and supplying a reason (a failed raise, a pivot, a departure) would be original research however plausible it read.

Reproducing a published method

Running someone else’s published method and reporting what it returns is description, provided the method, its inputs, and its limits are stated with the result. The DLT efficiency metric re-runs a 2023 comparison and publishes the 2026 figures next to the caveat that makes them honest: Radix is the only row whose search term changed between the two runs, so its two numbers are not measuring the same thing. The reading that caveat forecloses – treating the move from 123× to 51× as a gap closing – is the part that would have been original research.

A name is a claim about a person

The Ideas pipeline carries 29 cards tracking the Radix DAO transition, and 19 of them named an assignee – ten people, each rendered as a chip on the kanban board, as an Assignee row in the card’s own facts table, and returned to any agent that calls this wiki’s ideas tool. Every one of those names was derived the same way: from the author of the RadixTalk thread the card was built from. Opening a thread is a published, checkable act. Agreeing to deliver what it proposes is a second claim, and no source established it.

The gap between the two stopped being theoretical on 28 August 2026, when a member of the Radix DLT Official Telegram read this board as the only listing of open DAO tasks she could find – and one of the ten read it with her. Vlad asked whether he had been assigned tasks he knew nothing about, then said of the community-moderators card that he did not know where it came from, and of the website card that he had not dared commit to it while other work was outstanding. He had in fact opened both threads – Community Mods on 21 January 2026 and the website RFC on 19 February 2026 – and in the first he asked the channel for self-nominations rather than offering himself. The reading was right; the model laid over it was invented, the same shape as the quorum case above. On the website card the invention was visible from the card itself: its body already narrated a different contributor, Timan of Astrolescent, carrying a different and actually-voted proposal.

Two properties made it durable. The claim lived in a metadata field, which this wiki’s link and content checks do not read, so nothing audited it for the seven weeks it stood; and the person it described had no reason to look at a page about a task he had not taken. Both cards were corrected on the day, the field cleared and the row relabelled to what the thread actually shows: raised by, dated and linked. The rule the case leaves behind is narrow and absolute. An intention attributed to a named living person needs a source in which that person states it. Where the source establishes only that someone raised a subject, record that they raised it, and leave the ownership empty rather than filling it with the nearest name.

Where an inference belongs

A conclusion worth drawing is not worth suppressing; it is worth putting where it can be argued with. Analysis, forecasts, and interpretation belong in the Blog, where they carry a date and an author, or in the Ideas pipeline while a proposal is still live. Where someone else has published the analysis, an article may report it as theirs – attributed, dated, and not adopted into the wiki’s own voice, per neutral point of view.

Nor is the Ideas pipeline a place where sourcing stops mattering. It is the right home for a proposal that is still an argument, but a card is still published, still indexed, and increasingly still read as a record – and a claim it makes about a person is held to verifiability exactly as an article would be.

The same applies to the honest gap. Where a claim cannot be checked against the ledger, a filed document, or a published source – an off-ledger balance, an unannounced roadmap, a private agreement – write what is known, attribute it to whoever said it, and state plainly what is not established. Naming the gap is a stronger article than filling it.

See also

• Verifiability
• Neutral point of view
• Notability
• Freshness
• Conflict of interest
• Editorial notices