PubTech Radar Scan: Issue 33
Open infra blues, phishing alerts, launches (Medical Evidence Project, DocuMark, Preprint Watch, Vibe Coding, 2nd Opinion), AI search impacts, AI longer reads & insights from Aaron Tay.
It’s been a busy couple of weeks, with lots of announcements linked to the SSP meeting in Baltimore. Congratulations to all the Tools and Products award winners, many of whom have been highlighted in previous issues. I’m sure there’s much I’ve missed, so do drop me a line if there’s something you think I should highlight. There are plenty of longer AI-related reads in this issue, with a more reflective tone creeping in as we start to reckon with how AI might reshape publishing and scholarly communication.
News
🧱 The 2025 State of Open Infrastructure report paints a rather bleak picture of open infrastructures left to fend for themselves due to volatile funding environments. See also Jean-Sébastien Caux's open letter to the scientific community reveals SciPost's financial struggles, with only 11% of benefiting institutions providing support.
⚠️ Emerald Publishing is confronting a surge in sophisticated phishing scams that impersonate their brand and staff to deceive authors into paying for fake publication services.
🌀 Laura Harvey shares reflections from 25 conversations about current shifts in scholarly communication: omni-crisis is the new norm, AI’s shifting to infrastructure, B2C is booming, and AI fear is real.
🎤 In his keynote at Publisherspeak UK 2025, Ian Mulvany highlighted the critical role publishers play in shaping future knowledge systems through deliberate infrastructure choices. He stressed that decisions made now, especially regarding identity management, will have lasting impacts on the integrity and accessibility of scholarly communication.
Launches:
🕵️♂️ James Heathers’ Medical Evidence Project blends white hat hacking tactics with journalistic urgency, taking a direct route to expose flaws in medical studies. Rather than wait on lengthy investigations, he’s hacking the process, literally and figuratively, by taking findings straight to the press.
📘DocuMark, courtesy of Enago and Trinka AI, promises to rescue university staff from the whack-a-mole game of AI detection. DocuMark enables students to verify their AI usage, gives educators clearer insights into student authorship, and integrates into learning management systems. [Press release]
🚨Preprint Watch aims to spot paradigm-shifting research early. By classifying preprints based on their potential to disrupt, it offers a sort of early warning system for those in publishing, funding, or policy-making.
🖼️After years of quietly crafting open-source tools for publishing and the wider scientific community, eLife is now putting it all on display. Their new website highlights how they’re helping build open tech for open science.
🕵️ Molecular Connections has launched a confidential integrity-investigation service named "2nd Opinion". The service offers speedy access to a team of Subject Matter Experts and integrity specialists from Molecular Connections.
💻 Adam Hyde’s Vibe Coding Workshops are live!
AI & AI longer reads:
📚 David Worlock in AI and Publishing: Death, Revolution – or Oblivion?, notes that while publishers are understandably focused on AI's role in content creation, the more significant change may be how readers themselves are beginning to use AI to access information directly. I’m no longer convinced that disintermediation is the right framing. In academic publishing, I think we’re probably entering a phase of increased intermediation that will be less visible and shaped by Big Tech.
🔍“Google Zero”, a delightfully ominous term for when Google sends no traffic to publisher websites, is the hot topic following Google’s rollout of AI Mode to search. 📉 For more, see Sara Guaglione from Digiday on What Google’s new AI Mode means for publishers, try The Rebooting Show or Pete Pachal on AI Mode, “...singular, concise answers may have the inadvertent effect of flattening knowledge diversity. Mainstream perspectives will get amplified, and niche or contrarian voices will have a tougher time standing out.”
🔗In his latest Substack piece, Mike Caulfield explores the differences in link hallucination and source comprehension across large language models. Using his trusty SIFT Toolbox, he finds that models that fabricate fewer links tend to grasp their sources better, because, shockingly, not making things up does help with accuracy. ✅ More seriously, it’s a timely reminder that we need more benchmarking of this kind for academic contexts, and a growing suite of AI-wrapper tools to vet content not just for correctness, but for how well it aligns with sources.
🧠Rosalyn Metz examines the Wiley-Perplexity partnership, highlighting its potential to enhance academic content accessibility through AI-driven search. 💭I can’t help thinking that the future for both publishers and librarians will be somewhat bleaker. I don't think papers will remain the primary access point for academic STM knowledge. Instead, it will likely be through Big Tech platforms managed via the university rather than the library, with content filtered in ways many of us find uncomfortable but will accept for the sake of convenience. On more optimistic days, projects like the National Library of Norway’s national LLMs suggest a future that could still be more open, diverse, and library-led.
💡Chris Duncan floats a radical idea: what if news publishers joined forces to buy Chrome in a bold bid to wrest back control of content distribution and monetisation? But will we still need web browsers? 🕸️ Scott Rosenberg suggests we might be witnessing the browser's last hurrah, with browsers and search increasingly bypassing traditional clicks and content in favour of summaries and chat interfaces. This transformation, led by companies like Google, OpenAI, and newer players like The Browser Company, threatens to unravel the economic and creative incentives that kept the web vibrant in the first place.
🧨Anthropic CEO Dario Amodei recently warned that AI could wipe out up to 50% of entry-level white-collar jobs, possibly pushing unemployment to 10–20% in the next five years, causing considerable comment. 🧮 More hopeful takes come from Nathaniel Whittemore, reading and reflecting on Tim O’Reilly’s AI First Puts Humans First, and Jules White’s question: “Will companies use AI to help everyone do more—or just cut jobs and keep output the same?” White makes the case for the first option: “Empower your people. Expand your vision. Let AI scale your ambition, not shrink your workforce.” 💼 I’d like to believe that, but I think scholarly publishing will go the other way. The focus on efficiency and protecting profit margins makes a “do more with less” mindset hard to shake.
🧠 I learn so much from Aaron Tay and really appreciate the time he takes to put together his posts. Here are three great reads:
A Librarian’s Guide to AI in Academic Search Tools. A really good, but long, read that deserves your time.
How should academic retrieval augmented generation (RAG) systems handle retracted works?
Generative AI in Practice: EBSCO’s AI Insights by Richard Bridgen (found via Aaron's Bluesky feed). Interesting to see how this was tested and the conclusion: “Given the issues the testers raised to do with missing detail, inconsistency, the potential for bias and lack of customisation of the feature, it was decided not to switch it on until such time as these issues have been addressed.”
📺 Benedict Evans’s latest presentation, “AI eats the world”, is a sharp, accessible summary of where we are with AI today. Not publishing-specific, but a great contextual overview.
📋 In case you fancied a light 167-item browse, I’ve refreshed the AI-Powered Publishing Use Cases list.
And finally:
🦀 The Shawshank Crustacean. A lovely crash blossom: "The convicted killer crab walked between two walls and pushed through newly installed razor wire..."