The Quiet Revolution in Science Publishing: From Locked Journals to Open Everything

 

For centuries, the path from discovery to public knowledge ran through a narrow gate: the peer-reviewed journal, controlled by a handful of commercial publishers, funded by subscription fees or steep author charges, and often taking six to twelve months to let anyone outside academia see the result. That gate is being dismantled — not by a single revolution, but by a decade-long accumulation of tools, platforms, and models that are now converging with AI to make "free to publish, free to read" the default rather than the exception.

The recent launch of OpenSource: Cardiology — a nonprofit, physician-founded journal that charges nothing to authors or readers — is a good entry point into this story. But it's one data point in a much larger, already-mature movement that includes repositories like Zenodo, social platforms like ResearchGate, and a publishing philosophy called Diamond Open Access. Understanding how these pieces fit together — and where each one falls short — explains both why this tradition is spreading and why it hasn't replaced the old system.

The three layers of open science today

It helps to separate three things that often get lumped together as "open access," because they solve different problems and carry different guarantees.

Diamond Open Access journals (like OpenSource: Cardiology, JOSS, or eLife) are peer-reviewed publications that charge nothing to authors and nothing to readers, typically funded by philanthropy, institutions, or grants rather than Article Processing Charges. As of 2021 estimates, somewhere between 17,000 and 29,000 journals already operate this way, representing roughly 73% of journals listed in the Directory of Open Access Journals — meaning Diamond OA isn't a fringe idea, it's arguably the statistical majority of the OA landscape, even though it publishes a smaller share of total articles (about 44%) because these journals tend to be smaller and more specialized (Wikipedia: Diamond open access; Petite encyclopédie de l'édition scientifique ouverte).

Repositories like Zenodo solve a different problem: preserving and citing research objects — datasets, code, preprints, conference materials, negative results — that traditional journals often won't touch. Launched by CERN and OpenAIRE in 2013, Zenodo now handles millions of uploads and assigns a DOI to anything uploaded within seconds, making it permanently citable (SPARC: Zenodo's Open Repository). It doesn't peer-review anything; it's a trusted, EU-institution-backed vault, not a journal.

Networking platforms like ResearchGate solve the visibility and discovery problem: making it easy to find, follow, and message other researchers, and to get informal reads and citations on your papers. But this is also where the model gets shakiest — ResearchGate hosts both peer-reviewed and non-reviewed content side by side without clearly distinguishing them, and it's currently facing growing criticism over manipulated citation metrics and a lack of any formal vetting process for uploaded content (Nature: Does ResearchGate have a growing credibility problem?). A study of researcher behavior even found that using ResearchGate correlated with little measurable career benefit (Emerald: Social networking for scientists).

How the new tradition compares to the old one

Dimension	Traditional journals	Diamond OA journals	Repositories (Zenodo)	Networking sites (ResearchGate)
Cost to author	Often $0–$11,000+ APC or paywalled	$0	$0	$0
Cost to reader	Subscription/paywall common	$0	$0	Mostly free
Peer review	Established, but imperfect (see below)	Rigorous, but track record still building	None	Inconsistent/mixed with unreviewed content
Speed	Months to over a year	Days to weeks (AI-assisted triage)	Immediate (DOI in seconds)	Immediate
Indexing/prestige	PubMed, Web of Science, high impact factor	Often lacks indexing initially	Not indexed as a "publication"	Not a formal publication venue
Funding model	Subscriptions, APCs	Philanthropy, grants, modest non-research fees	Public/EU institutional funding	Freemium/ad-adjacent
Best for	Career-critical, high-stakes findings	Original research authors wanting free, fast, ethical publishing	Data, code, negative results, priority claims	Discovery, informal feedback, networking

None of these models is strictly "better" — they're increasingly complementary layers of a single pipeline. A researcher might post a dataset on Zenodo for a DOI and priority claim, share a preprint for early feedback, network the results on ResearchGate, and then submit the polished manuscript to a Diamond OA journal for the peer-reviewed version of record.

What openness adds — and where it fails

The honest case for openness rests on speed and equity. Preprint culture — the direct ancestor of the Zenodo/ResearchGate ethos — demonstrably compresses a 6–12 month review cycle down to days, lets researchers in under-resourced countries read cutting-edge work without a subscription, and creates a timestamped record of priority that protects ideas from being scooped (Cureus: Pros and Cons of Preprints; Editage: What are preprints?). During COVID-19, this speed had tangible public-health value.

The honest case against unchecked openness is real too: without any review, flawed methodology can circulate freely, and journalists sometimes report preprint findings as settled fact, only for conclusions to shift later. ResearchGate's blending of reviewed and unreviewed content in the same interface makes this worse by obscuring which is which (Nature).

Peer review's own credibility problem

It would be misleading, though, to frame this as "sloppy open platforms vs. rigorous peer-reviewed journals." The traditional gatekeeping system has its own well-documented failure modes — and recent research suggests they're getting worse, not better.

• Retractions are rising, not falling. The Retraction Watch database now tracks more than 65,000 retracted papers, and the retraction rate has climbed for two decades — these are papers that passed peer review before being pulled (Retraction Watch; ScienceDirect: retracted publications; PMC: The Secret Life of Retractions).
• Peer review often doesn't catch the errors it's supposed to catch. In a controlled experiment, reviewers given a paper with eight deliberately planted errors spotted only two on average, and a widely cited review concluded peer review's effects are "largely untested" (disabled-world: Peer Review, A Flawed Scientific Publishing Process; PMC: Peer Review in Scientific Publications).
• Paper mills have industrialized fraud inside peer-reviewed journals. A 2025 machine-learning audit of the oncology literature from 1999–2024 found that nearly 10% of the entire published, peer-reviewed body of work carried the hallmarks of paper-mill fabrication — and the fraudulent share has been rising in recent years (OncLive: Concerns Escalate About the Fundamental Integrity of Scientific Peer-Reviewed Literature).
• The system is straining under its own logistics. Reviewer fatigue and shortages since COVID mean papers increasingly get rejected or delayed for reasons unrelated to quality; submission-to-publication timelines in some fields now stretch past 300 days on average (Duck of Minerva: Is Peer Review Broken?; Scholarly Kitchen: Peer Review Has Lost Its Human Face; PMC: When peer review drags on).

None of this means peer review is worthless — it still catches plenty of bad science before publication. But the fair comparison isn't "flawed openness vs. flawless gatekeeping." It's two different, both-imperfect systems: one that trades speed for a risk of premature, unvetted claims, and one that trades speed for a slower process that still lets fraud, error, and bias through at a measurable and apparently growing rate.

Where AI-native journals fit into this

What's new in 2026 isn't the Diamond OA funding model itself — that's been documented for over a decade — but the fact that AI has collapsed the engineering barrier to launching one. Building a submission portal, a review-routing system, and an integrity-checking pipeline used to require a publishing company's infrastructure; now two practicing cardiologists can build it themselves, as Saxon and Sperry did with OpenSource: Cardiology. This is why you're seeing a wave of specialty-specific, physician-led Diamond OA journals emerge rather than just a handful of large, foundation-backed ones.

The tradeoff those journals face is the same one Zenodo and ResearchGate never fully solved: credibility takes time to earn. A new Diamond OA journal without PubMed indexing is in the same trust-building position ResearchGate was in twenty years ago — it has to prove, upload by upload or issue by issue, that "free and fast" doesn't mean "unreliable."

A proposal for a single rulebook across formats

If retraction rates keep climbing inside "gold-standard" journals while open platforms keep multiplying outside them, the deeper problem isn't which venue to trust — it's that trust is still assigned by venue rather than by conduct. This is precisely the gap addressed by the Memorandum on Personalized Responsibility, Unified Oversight, and Human–Artificial Participation in the Global Scientific Literature Space, published in July 2026.

The memorandum's central claim fits everything above: journals, preprint servers, repositories, social platforms, and AI-mediated systems are no longer separate boxes — they're one unified real-virtual literature space, and pretending otherwise is why the same fabricated result looks "trustworthy" in a journal and "unreliable" on a blog, when the actual methods and data are identical. Its core proposals include:

• A "scientific passport" — accountability tied to identifiable persons and their conduct (authors, reviewers, editors), not to the prestige of the platform they published on, with protections allowing identities to be held under seal in high-risk contexts so whistleblowers and vulnerable researchers aren't exposed to retaliation.
• Peer review recognized as citable scientific work in its own right, rather than invisible, uncompensated labor — reviews would be attributable where safe, and structured but accountable where anonymity is still needed.
• A clear line between honest disagreement and deliberate fraud: contested findings should be met with visible, citable counter-evidence rather than suppression, while fabrication and manipulation should trigger consequences tied to the person responsible, regardless of whether it happened in a top journal or an open platform.
• Federated, cross-disciplinary arbitration bodies to apply shared integrity standards across the whole literature space, since no single journal or platform can govern an ecosystem this large alone.
• AI treated as both a governed object and a governance tool — its role and training must be disclosed when used in research or review, while it can simultaneously help detect fraud networks and manipulation patterns, with final judgment always resting with identifiable human stewards.

Given the retraction and paper-mill data above, this reframing looks less like idealism and more like a direct response to the numbers: a prestigious venue and an open platform are equally capable of carrying dangerous fabrication, so the rules protecting readers should follow the people and the evidence, not the letterhead. The authors describe it explicitly as a working, open invitation for critique rather than a finished doctrine — full text available via ResearchGate and the author's blog.

The likely trajectory

Rather than one model replacing another, the evidence points toward publishing power itself shifting: away from a small set of institutional gatekeepers and toward distributed networks of accountable individuals  — researchers, reviewers, and readers acting more like a self-governing community than passive subjects of a handful of publishing houses. Diamond OA journals, Zenodo-style repositories, decentralized peer-review experiments, and the memorandum's scientific-passport idea are all expressions of this same democratic wave, each attacking a different layer of the old gatekeeping monopoly — access, evaluation, and accountability. Traditional subscription and hybrid-APC journals will likely persist for the highest-prestige work, but the deeper shift is structural: if integrity keeps depending on venue rather than on traceable, accountable conduct, the retraction and fraud numbers suggest the old system's "friction" will keep failing quietly, in exactly the venues it was supposed to protect — and whether this democratization makes science more trustworthy or simply noisier will depend on how fast verification infrastructure like federated arbitration and governed AI can keep pace with it.

You can find the full text of the memorandum here:

https://www.researchgate.net/publication/408350217_Memorandum_on_Personalized_Responsibility_Unified_Oversight_and_Human-Artificial_Participation_in_the_Global_Scientific_Literature_Space

and here

https://yabluchanskiy.blogspot.com/2026/07/memorandum-on-personalized.html