Decentralized publishing and peer review is the DeSci pillar that targets how research is disseminated and validated, sitting alongside novel funding, IP-NFTs, and DeSci DAOs. The traditional stack – a handful of commercial journals gate-keeping both distribution and validation – is the part of science the Ethereum Foundation's DeSci hub singles out for “creating, reviewing, crediting, storing, and disseminating” knowledge on open infrastructure.
The core move is to publish a research object rather than a document: manuscript, data, code, figures, and metadata bundled as one versioned, independently-addressable unit.
ResearchHub is the most-cited working model for the paid-review thesis, and in mid-2026 it narrowed sharply. Through at least 7 July 2026 its journal fronted a general submission route – “Publish Faster. 14 days to peer review. Immediate preprints. Open access by default,” with a Submit Your Manuscript button. By the 7 August 2026 capture that was gone. The journal (ISSN 3070-3395) now “exclusively publishes Registered Reports that originate from proposals funded on ResearchHub”: hypotheses and methods are locked in and published before data collection, results follow afterwards whatever they show, and every article chains back to the funding opportunity that paid for it. There is no article-processing charge and no route in for work ResearchHub did not fund.
Two numbers this page previously carried did not survive that re-scope, and both are worth restating precisely. The current editorial and peer-review process sets no per-review rate at all – reviews are still described as paid, but no figure appears in any current journal document. And the fourteen days is the window a reviewer has to return a report after agreeing to review, not a publication turnaround: the stated target for an editorial decision is 21 days from submission, against a minimum of two openly-published reports per version. RSC still funds bounties across the platform, though its governance role is thinner than it looks – RSC has no on-chain voting machinery, and Snapshot votes are non-binding signals.
The token moved onto major venues over 2025 – listed on Coinbase (Base network) on 31 July 2025 and Kraken on 17 November 2025 – and ResearchHub has integrated with Bio Protocol and Molecule to route RSC bounties into an on-chain research workflow, pulling peer review directly into the BioDAO funding pipeline.
Under the platforms sits open, publicly-inspectable infrastructure. DeSci Labs' Nodes are open-source “secure PID containers for interoperable research objects,” and the dPID protocol standardizes how those objects are named and resolved so citations point at durable, content-addressed artifacts rather than link-rotting URLs. Because objects are versioned and fingerprinted, the provenance chain – who published what, when, and what changed – is auditable rather than taken on trust. This is the same shift the rest of DeSci makes with IP-NFTs: turn an opaque institutional record into an open, portable, on-chain one.
A page about durable citation should be willing to check one. Crossref is the registry that mints and resolves the DOIs almost every journal cites itself by, and its API is keyless, so the record a publisher has actually deposited is a primary source anyone can read. ResearchHub deposits as member 33940, and on 19 August 2026 that member carried 82,383 registered DOIs. Enumerating all of them separates two namespaces under the 10.55277 prefix: 82,098 platform DOIs of the form researchhub.<id>, minted for papers and posts on the site, and 284 journal DOIs of the form rhj.<id>. The 284 cover 235 distinct article identifiers; the balance are versions, running as high as .7.
Three things in that record are worth stating plainly, because none of them is visible from the journal's front page.
rhj record carries a container title or an ISSN, so the journal's own identifier does not gather its own articles: api.crossref.org/journals/3070-3395 returns 404, and a works query filtered on that ISSN returns zero results. The articles exist and resolve individually; as a journal, in the index most citation tooling reads, it is not there.test or fake paper marker – leaves 77. The remainder are placeholders registered as live DOIs: asdfasdf, sdfsdf, This is a test paper. They are not inert records. 10.55277/rhj.k25xmc37, titled asdfasdf, redirects to a live ResearchHub page and returns 200.staging.researchhub.com, a non-production host that answers HTTP 500. Five of those sit inside the 77 that pass the title filter, so this is not confined to the obvious test rows.The registration dates for those 77 give the journal's real cadence: 13 in May 2025, 7 in June, 15 in July, none at all between August 2025 and February 2026, then 31 in March 2026, 6 in April, 3 in May, none in June, 2 in July, and none in the first nineteen days of August. Read alongside the re-scope above, the shape is coherent rather than damning – a general-submission journal that ran thin, and has been rebuilt around the one input its parent platform can guarantee.
On 2 October 2026 the member still reported 82,383 DOIs, a works query for anything created after 5 August returned zero, and api.crossref.org/journals/3070-3395 still 404s. Re-read on 27 August 2026, every figure above still holds: 82,383 registered DOIs, 82,098 researchhub and 284 rhj, 235 distinct journal identifiers, no container title or ISSN on any of them (api.crossref.org/journals/3070-3395 still 404s, and a works query filtered on the ISSN still returns zero), and the same 32 journal DOIs dereferencing to staging.researchhub.com. What the earlier pass did not measure is the cadence of the other namespace, and it is the more revealing of the two. The 82,098 platform DOIs are not a steady deposit.
Across the forty-five months from March 2022 to November 2025 the member registered 1,147 of them; then 80,884 – ninety-eight per cent of the namespace – arrived in six months: 416 in December 2025, 532 in January 2026, 6,481 in February, 29,144 in March, 30,837 in April and 13,474 in May. Since 1 June 2026 the whole member has registered 67 DOIs, the most recent of any kind on 5 August 2026, and the journal namespace has been silent since 30 July. A prefix that emitted thirty thousand identifiers in a month and three in the latest one is recording a retrospective backfill, not a publishing rate – which matters to anyone reading 82,383 as a measure of output. Counts derived by enumerating the member's full works list through the keyless Crossref REST API.
The wider point is methodological, and it cuts against the sector's own marketing. Decentralized publishing sells itself on identifiers that do not rot, and the sales pitch is usually made in prose. The identifier record is checkable in a way the prose is not, it is public, it costs one unauthenticated request, and it disagrees with the prose more often than anyone checks. That applies to dPID and Codex exactly as much as it applies to Crossref.
Caper is a DAO protocol, not a publishing platform – it does not host papers or run peer review. But it shares the load-bearing principle underneath decentralized review: credit and influence should be earned and non-transferable. Paid on-chain peer review works precisely because a reviewer's reputation is accrued by contribution and stays attached to them; it is not a badge anyone can buy.
Caper's governance carries the same idea into who steers a treasury. A member earns votes only by casting ranked ballots on the caper's proposals; buying and selling the caper's token earns none, and the votes sit in a soulbound token that cannot be sold or handed on. A member's say, and their share of the treasury when they exit, grow with both those votes and the tokens they hold.
That record is the property the reviewer-reputation model in on-chain publishing is trying to build: it cannot be bought on the market, and the only way to earn it is to turn up and vote. Holdings still count, so a caper is not run without capital, but a big bag alone can't capture control.