Understanding Open Access Publishing

⚠️💰 134 500 000 € for €lsevier from France 💸 France has just signed a "transformative agreement" with Elsevier https://lnkd.in/eaSr2PRh, and I believe this deserves our collective attention. As an Open Science enthusiast and co-founder of JTCAM: Journal of Theoretical, Computational and Applied Mechanics, I'd like to share some critical insights: 📊 The Agreement: Cost: 😮 €134.5 million over 4 years Covers 241 French institutions (44 of which are free of charge) Approximately €690,000 per paying institution 💰 To put this in perspective, compare it to annual public funding for various French institutions and organizations: + CCSD (HAL, ...): 0.4 M€ + Génopole - 7.3 M€ + ENSTA ParisTech - 20 M€ + Ecole des ponts - 32 M€ + ISAE - 44 M€ + Ecole polytechnique - 110 M€ + INRIA: 216 M€ CCSD with its HAL, Episciences and ScienceConf projects is a leading Open Science actor in France and Europe and its annual budget is only about 1% about 1% of the annual cost of Elsevier! 🤔 Critical Questions: + Why are we paying this exorbitant price for Open Access (OA) publishing with Elsevier while we have all the necessary infrastructure (HAL) and a will to share them without extra cost and an embargo period? + How will this agreement impact the future of scientific publishing and its costs? 🔍 Key Points: + France already has the necessary infrastructure for OA through HAL, without additional costs or embargoes thanks to the rights retention strategy adopted by France. + Despite opposition from some institutions (e.g., University of Lorraine), Elsevier has succeeded in selling its Gold OA under the guise of a "transformative agreement". + This is disappointing for Open Science, as France could have made all publications open without this agreement and its additional costs. + We could find a better use of this extra OA fee integrated in this contract, e.g. spend it on research funding and personnel. ⚠️ Hidden Costs: While we won't see direct OA publication fees as labs or researchers, these costs still exist. Be cautious when publishing OA articles, as the volume of these will likely influence negotiations in 4 years. 🌟 My Stance: I will not publish OA with Elsevier under this agreement. Instead, I encourage utilizing existing national infrastructures like HAL for true open access and publishing in Diamond Open Access journals like, for example, CR Mécanique and JTCAM: Journal of Theoretical, Computational and Applied Mechanics. 👩🔬 👨🔬 👩🏫 👨🏫 👩💻 👨💻 What are your thoughts? How will this agreement impact your publishing practices? Let's discuss the implications for the global scientific community. 🔗 For more information: Systematic analysis of these transforming agreements https://lnkd.in/eEeEGfB4 #OpenScience #ScientificPublishing #Research #TransformativeAgreements #AcademicPublishing #JTCAM

Open access ≠ high publishing fees Cancer Research UK has recently announced it will stop funding open-access publishing costs from 1 October 2026. But it will still require CRUK-funded research papers to be made openly available. Similar decisions are being made in other countries, too. To me, this is a much bigger signal than one policy change. More funders may start asking a simple question: Should scarce research funds be spent on publication fees or on the research itself? Because these are not the same thing. It's important to emphasise that this is not anti-open-access. CRUK still expects funded papers to be openly available, and its policy still allows routes such as green open access, Europe PMC deposit, Octopus, self-archiving and institutional agreements. That is the real point: We should separate “making research open” from “paying whatever the publisher charges”. There are many ways to make research more open and more useful: - repository deposit - preprints - white papers - webinars - lower-cost publishing routes Preprints are still encouraged in CRUK’s policy, but the final peer-reviewed paper must also meet the open-access requirements. My view? If a fixed pot of money exists, I would rather see more of it spent on: - better research, - better researchers, - better collaboration, - and better dissemination Not just higher publication fees. The future of open research should not depend only on who can afford the APC. It should depend on how well we build systems that make knowledge accessible. Open access matters. But we should be honest: APCs are only one route to openness. And often not the best one. #publishing #science #research #researcher #phd #postgraduate #apc #journal #professor #academic #nature

The four largest commercial publishers all have open access growth baked into their forward guidance. Springer Nature launched a new fully OA Nature Progress journal series in March 2026, with the CEO framing growth as "driven by our continued leadership in open access". Elsevier's parent RELX expects continued strong STM growth, with pay-to-publish named as the primary research growth driver. Taylor & Francis reports open research volumes "continue to grow strongly", on the back of 20%+ submission growth last year. Wiley says its author-funded OA is "growing consistently above 20%". I find this confident bet on continued expansion increasingly hard to square with what funders are doing across multiple jurisdictions: - In the UK, UKRI confirmed last week it will stop funding hybrid OA charges from 2028/29. Cancer Research UK goes further: all funding for article publication charges ends October 2026. - In the US, the 2027 White House budget request asks Congress to prohibit federal funds for "high-cost subscriptions and APCs". The National Institutes of Health has recently closed a consultation on capping APCs; responses are published but no decision announced so far. - In China, the Chinese Academy of Sciences, the world's largest research institution, with 50,000+ researchers, is reportedly halting payments for journals charging above $5,000. - In Switzerland, where 80% of SNSF-funded research is already open access, the SNSF will from January 2027 cap APCs at CHF 3,500. It is also funding Diamond OA via Open Research Europe, suggesting some of the savings from the cap will be redirected into alternative models. Publishers typically offer earnings guidance only a year ahead, and several of these funder changes don't bite until 2027 or 2028. STM publishing is also remarkably resilient and resistant to disruption. All the same, it looks to me as if the publicly announced policy shifts aren't yet being priced into the growth thesis. Researchers will keep submitting and volumes are likely to keep growing, but it's increasingly unclear who will pay, at what level, and for how long to make this content OA. Cash-strapped libraries are unlikely to take up the slack as funders step back. Which leaves the harder question: will green and diamond OA finally come to the fore - or are we heading back to subscriptions by another name? #OpenAccess #ScholarlyCommunication #ResearchPolicy #AcademicPublishing #ResearchFunding

The transition to open access appears to have stalled. Why? Yesterday I posted an essay that explores the link between the rapid growth of research articles from China and the proportion of papers published open access globally. The attached document summarises the key takeaways. To get the full story (and the caveats) you can read the essay here: https://lnkd.in/evvVtDUV The scholarly publishing industry is going through a significant transition. We've never experienced article growth like this before, driven by a country that owns relatively few journals that its researchers publish in. Historically, Chinese institutions have paid less under 'read' subscription models than institutions in the USA, Japan and Europe. This makes transformative agreements hard to negotiate, as Chinese institutions would need to pay a lot more than they do now. (And no one wants to pay more.) There are many questions with unknown answers: ➡️ Will western publishers push back on publishing Chinese research if they can't monetise those papers adequately? ➡️ Will China be willing to pay more than it does currently to support scholarly publishing infrastructure? ➡️ Will Chinese funders mandate their researchers to publish more content in new, local journals? Or will they prefer to publish in established journals for free under a subscription reader-pays model? ➡️ What can be done to increase the peer reviewer pool in China? These are big, difficult questions for a Saturday morning. Thoughts welcome!

I edit manuscripts for Springer Nature. Read posts from Jonas Heller and Alex Zhavoronkov about the topic: The future of the publishing system. Last month, an AI-generated paper passed peer review at an ICLR workshop. Cost to produce: $15. Time: a few hours. The reviewers couldn't tell. Neither can I, sometimes. Here's what most people miss about the "crisis" in scientific publishing: It's not a crisis. It's a phase transition. NeurIPS submissions: 3,297 in 2017. 27,000 in 2025. Reviewer acceptance rate: collapsing. Papers with paper mill signatures: 1 in 50. LLM fingerprints in peer reviews: up to 17%. The system isn't stressed. It's breaking. I send reviewer invitations every week. Decline. Decline. No response. Decline. Scientists contribute 130 million hours annually to peer review. Zero compensation. Zero training. Zero recognition. Why would they keep doing it? Meanwhile, in laboratories: Pacific Northwest National Lab just deployed autonomous research platforms. Robots run experiments 24/7. AI agents design protocols, analyze data, adjust parameters. Lawrence Berkeley's A-Lab synthesized 41 novel materials autonomously. Sakana's AI Scientist writes complete research papers for $15. The endpoint is obvious: AI generates hypotheses. Robots run experiments. AI writes papers. AI reviews papers. Where exactly do humans fit? My prediction: the current system survives five years. Maximum. Not ten. Five. The economic logic is unforgiving. What changes: Phase 1 (now): Detection arms race. Tortured phrase databases. Prompt injection attacks. Fraud detection theater. Phase 2 (2027+): The literature bifurcates. High-trust venues become invitation-only. AI-generated content floods open-access mega-journals. Phase 3 (2029+): Validation becomes the product. Reproducing findings matters more than publishing them. Autonomous labs become arbiters of truth. If your value is "I write papers," you're competing with $15 software. If your value is "I interpret what matters and make ethical judgments about risk," you're still ahead. The uncomfortable truth: academic incentive structures still reward publication volume. The metric is becoming meaningless. The careers built on it are not. I don't have a clean solution. But I know what doesn't work: Pretending the system is fine. Defending what's already failing. Optimizing for metrics that AI can generate at infinite scale. The question isn't AI versus humans in science. The question is: What do we want the scientific literature to actually do? Archive everything? AI wins. Curate what matters? Different system required. Validate truth claims? Reproducibility infrastructure, not publication infrastructure. We can't have all three. Not anymore. I still review manuscripts. I still believe expertise matters. But I'm planning for what comes next. You should be too. Full analysis linked below ⬇️ What's your timeline? Am I too pessimistic? Or too optimistic?

Let’s talk about open science. A scientific breakthrough reaches its full potential when it empowers the broad community to replicate and expand upon findings. Over the last decade, we have developed and maintained several key technologies and datasets empowering an active ecosystem of more than 250,000 researchers and developers worldwide. Our commitment to responsible open science spans several critical frontiers: 🧬 Genomics: Our deep learning tools—DeepVariant, DeepConsensus, and DeepPolisher—have enabled the global community to process the exomes and genomes of 2.5 million individuals. 🧠 Neuroscience: We provide technologies like Neuroglancer and TensorStore alongside massive public datasets, such as the 1.4 petabyte H01 sample of human brain tissue. 🌍 Earth AI:  We are providing critical data for climate adaptation and urban development, including 1.8 billion detections in Open Buildings and the NeuralGCM atmospheric model. We’ve also released the Groundsource dataset for urban flash floods, comprising 2.6 million historical flood events derived using Gemini on 20 years of public data and spanning more than 150 countries. And Caravan: is a community-driven dataset for large-sample hydrology. 🏥 Healthcare: Our Health AI Developer Foundations (HAI-DEF) and Open Health Stack are democratizing medical AI, with deployments in over 10 countries reaching 65 million beneficiaries. 🦒 Biodiversity: SpeciesNet, a global-scale model, is currently classifying 2,498 animal categories to support conservation efforts. The real-world impact is the ultimate goal. From predicting floods in 150 countries covering 2B people, to predicting monsoons for 38 million farmers in India to preserving the critically endangered kākāpō in New Zealand, open resources allow for faster research and deployment As we transition into the era of AI-enabled science, we believe agentic workflows will allow scientists to transform complex methodologies into accessible, scalable tools. In this fast-paced new paradigm, communication and collaboration are more critical than ever. Open-source software and open datasets serve as the essential foundation for this ecosystem, enabling faster innovation and universal sharing of scientific knowledge. Read about real world impact powered by open science in today’s blog by our science team: goo.gle/4dnd9hx

Public health funds the science. Publishing is pricing it out. Across public health globally, epidemiologists are expected to: → Generate evidence for policy → Respond to outbreaks → Publish rigorously → Share findings quickly But here’s the uncomfortable reality: To publish open access in many top journals today, public health researchers often must pay $6,000–$13,000 per paper. At the same time: → Reviewers work for free → Editors are largely unpaid academics → Grants are capped → Public health agencies and universities rarely have dedicated APC budgets. This isn’t a niche problem. It directly affects who gets to contribute evidence. → Early-career researchers → Public health practitioners → Global collaborators → Researchers outside elite institutions They are quietly filtered out. Swiss universities recently took a public stance against this model. The message was clear: open science cannot survive on closed economics. In epidemiology, this matters even more. Our data are public. Our work informs policy. Our findings affect lives. If only well-funded institutions can publish, then evidence itself becomes biased. This isn’t about prestige journals. It’s about whether public health research remains truly public. I hope the global academic and funding ecosystem is paying attention. #PublicHealth #Epidemiology #OpenScience #ResearchEquity #AcademicPublishing

The airwaves are buzzing with AI for science and drug discovery, and co-folding is a big part of that momentum. Progress on open co-folding platforms has attracted enormous visibility. Visibility attracts investment. And too often, the next step follows a familiar script: capabilities that began open become increasingly closed (e.g. AlphaFold, Boltz, Chai, and others). That may be rational for individual companies and investors. But for the broader scientific community, IMHO, it can be a real problem. Open science provides critical infrastructure, tools, data, and methods whose impact extends far beyond the projects that create them. If we want open solutions to important, high-value problems in structural biology and drug discovery, we have to build, sustain, and support them intentionally. Openness is not guaranteed. And the value of open science extends far beyond the projects that create it. A great example is OpenProteinSet, a massive open-source dataset created by the OpenFold team to provide the training data needed to build, fine-tune, and evaluate deep learning models for structural biology. That infrastructure has helped enable subsequent efforts across the field, including Boltz, Chai, Protenix, and more recent work such as OpenDDE. This is the compounding value of open infrastructure. The code, weights, models, and datasets released openly today become building blocks for innovations we cannot yet anticipate, including commercial ones. Open science is not the opposite of innovation. It is often the foundation that makes innovation possible. As AI for science becomes more capable and more valuable, I believe maintaining genuinely open data, infrastructure, and foundation models will become increasingly important. We need models the community can inspect, reproduce, fine-tune, improve, and build upon. That is why efforts like the OpenFold Consortium matter so much. The future of scientific AI will not remain open by default. We have to choose to make it open. Choose wisely. #OpenScience #OpenFold #AIforScience #DrugDiscovery #StructuralBiology #Cofolding

The Bill & Melinda Gates Foundation will now require grant recipients to share their research manuscripts as preprints, the first major funder to take such step. They will also stop covering fees for open access publishing. I spoke with researchers and publishers on how these changes might impact publishing. On the one hand, more research will become freely available in preprint. On the other, the final published versions of articles, known as the version of record, might become harder to access, says Lisa Janicke Hinchliffe, a librarian and academic at the University of Illinois Urbana–Champaign. Ending support for article-processing charges is a “very sensible plan” given the unsustainable increase of such charges in recent years, says Shina Caroline Lynn Kamerlin, a computational biophysicist at the Georgia Institute of Technology in Atlanta. To learn more, read my story at Nature (Nature Portfolio): https://lnkd.in/dCrKM5q6

🎓 3 Papers Accepted in the last few days, and why I’m feeling conflicted. I’m proud to share that our team just had three papers accepted for publication. It’s always a milestone to see hard work recognized by traditional journals. However, I’m finding it harder to reconcile the way we are evaluated with how science actually moves today. Current academic metrics still rely heavily on the "journal stamp," a process that often takes months (or years). Meanwhile, I want to see us transition to a system focused on immediate preprint curation and AI-driven validation. Instead of waiting for legacy benchmarks, we should be leveraging tools that can evaluate research quality the moment it’s ready. To walk the talk, I’ve decided to share the AI review reports from PeerAI for these papers rather than the final journal links. Why PeerAI? Speed: Getting an objective, comprehensive evaluation in minutes, not months. Transparency: Focusing on the raw data and methodology through automated, unbiased analysis. Future-Proofing: Shifting the focus from "where" a paper is published to the actual "merit" detected within the preprint. The Reality of the "Untenured" Path I’ll be honest: as an untenured scientist, I’m still operating within the system. For the near future, I will continue to submit to journals and perform reviews for those I deem non-predatory. My career—and the careers of those I hire—currently depends on it. But my hope is that we reach a point where we can stop using journals entirely. This is a transition I’m navigating carefully, and it’s a conversation I plan to have individually with every person who joins my team. I believe the future of research curation is automated, fast, and open. Let’s start evaluating work based on the science itself, right when it’s released. Watch for posts sharing the PeerAI review reports soon. I'll also highlight which specific parts we changed in the final versions that match peerAI feedback. #OpenScience #AI #Preprints #AcademicInnovation #PeerAI #ResearchImpact #AcademicLife