J’ai mis en libre accès un outil que j’utilise au quotidien : un skill Claude pour faire une revue de littérature rapide sur PubMed.

Le principe : vous posez une question clinique précise, par exemple « quelle prise en charge pour une pseudarthrose du scaphoïde chez l’adulte jeune ? ». L’outil procède ainsi :
- il reformule la question en PICO et propose un plan de recherche ;
- il lance une série de recherches PubMed (5, 10 ou 20, au choix), en affinant les termes à chaque tour ;
- il vérifie que chaque référence citée existe bien dans les résultats PubMed de la session ;
- il livre un document Word : une synthèse, un tableau des preuves et une liste d’articles à lire, avec les liens DOI et PubMed.
Il fonctionne grâce au connecteur MCP PubMed de Claude.
Les recherches portent donc sur la vraie base PubMed, pas sur la mémoire du modèle.
Ce que ce n’est pas : une revue systématique. Il n’y a ni double sélection, ni évaluation du risque de biais, ni méta-analyse.
C’est une carte de la littérature pour s’orienter vite et savoir quoi lire en premier. Il n’a pas été validé à grande échelle : il fait ce qu’il annonce dans mon usage, rien de plus. Chaque référence doit être relue avant d’être citée.
L’idée originale vient de Consensus.app. Je l’ai adaptée à PubMed et à la pratique clinique.
Le skill est gratuit, sous licence MIT.
Les retours, critiques sont les bienvenus.

https://lnkd.in/e5kwd3_7

#ChirurgieDeLaMain #RechercheClinique #IA #PubMed #OpenSource

_____


I’ve made a tool I use every day freely available: a Claude skill for fast literature reviews on PubMed.

How it works: you ask a precise clinical question, for example “how should scaphoid nonunion be managed in young adults?”. The tool then:
- reframes the question as PICO and proposes a search plan;
- runs a series of PubMed searches (5, 10 or 20, your choice), refining the terms each round;
- checks that every cited reference actually appears in the session’s PubMed results;
- delivers a Word document: a synthesis, an evidence table and a reading list, with DOI and PubMed links.
It runs on Claude’s PubMed MCP connector, so the searches hit the real PubMed database, not the model’s memory.
What it is not: a systematic review. There is no dual screening, no risk-of-bias assessment and no meta-analysis. It is a map of the literature to get oriented fast and know what to read first. It has not been validated at scale: it does what it says in my own use, nothing more. Check every reference before you cite it.
The original idea comes from Consensus.app. I adapted it to PubMed and clinical practice.
The skill is free, under the MIT license.
Feedback, criticism are welcome.

https://lnkd.in/e5kwd3_7

#HandSurgery #ClinicalResearch #AI #PubMed #OpenSource

Does anyone at BMJ know how I can submit to their Christmas issue? I have an anlaysis looking at citations ranking papers/authors by how funny they are..

I call it the h-ha index!

(Serious tho)

10 AI Tools Every Academic Researcher Should Know About in 2026! 👇
AI is transforming academic research, making literature exploration, paper analysis, writing, and evidence synthesis more efficient.
Here are 10 practically useful AI tools for researchers and PhD scholars:
1. ChatGPT
➤ Brainstorming and research questions
➤ Conceptual explanations and paper structuring
➤ Draft refinement, coding and data interpretation
2. Claude
➤ Long-document and thesis analysis
➤ PDF reading and manuscript feedback
➤ Academic writing support
3. Perplexity AI
➤ Literature exploration
➤ Source-backed answers
➤ Recent research and topic scoping
4. Elicit
➤ Literature reviews
➤ Research data extraction
➤ Evidence synthesis tables
5. Consensus
➤ Evidence-based academic searches
➤ Answers from peer-reviewed studies
➤ Exploring scientific consensus
6. SciSpace
➤ Understanding complex papers
➤ PDF interaction
➤ Explaining research methods and findings
7. ResearchRabbit
➤ Discovering related papers
➤ Citation networks
➤ Mapping authors and research connections
8. Litmaps
➤ Visual literature mapping
➤ Tracking research developments
➤ Identifying overlooked studies
9. scite
➤ Analysing citation contexts
➤ Identifying supporting or contrasting evidence
➤ Evaluating research claims
10. NotebookLM
➤ Organising research documents
➤ Source-grounded summaries
➤ Research notes and study guides
💡 A word of caution: AI is a research assistant, not a substitute for scholarly judgement. Always verify references, validate findings, assess evidence critically, and maintain academic integrity.
The real value of AI lies not in using more tools, but in using the right tools intelligently.
Which AI tool has made the biggest difference to your research?
#AcademicResearch #AIinResearch #PhDResearch #ResearchMethodology #HigherEducation #JGU # ODS

It's important to understand that the Open Source Initiative (OSI) never worked to define open source (maybe a bit with AI), but only to protect the "Open Source Definition" — both are incompatible. They missed decades of debates regarding the meaning of open source beyond software.

- 2006: in a critics of Creative Commons Non-Commercial/derivative clauses, the Definition of Free Cultural works was suggested. In order to guarantee freedom they mention the need to access to source files.
[https://lnkd.in/ddHgzjAA]

- 2007: writing of a research article "Defining Open Source" by Russ Nelson, former ephemere OSI director (in 2005), with a dedicated "beyond the Definition" mentionning that "The concept is applicable to anything which has separable design elements."
[https://lnkd.in/dBJK3h23]

- 2014: A github employee, Ben Balter (in interaction with Karl Fogel), while working on the glossary wrote the article "Open source, not just software anymore" and extended the notion to cover other kind of editable resources. The definition of the main open software platform is not software centric. « @benbalter The internet has noticed that you tried to define “open source” », but the organisation that supposedly define it didn't.
[https://lnkd.in/dH_WCPed]

- 2017: Launch of the journal of Open Source Education (JOSE) section, an "an affiliate of the Open Source Inititative". In their about section (that can be find from first availble archived version) they explain the notion of "open-source educational materials". « "Open-source" implies that there are source files that are converted and rendered into a final learner-presentable form. »
[https://lnkd.in/dz9a_CAy]

- 2021: paper "Open Source Educational Resources" to distinguish OER with available source files, open source being theorized explictely beyond software, echoing (JOSE) position.
[https://lnkd.in/dHCE-_Ji]

- 2023: I asked the OSI directly by email around "open source image", a PNG with available svg file. An OSI staff told me "If you provide the source for the image so that I can modify it and it uses an OSI compliant license, then yes, I would say it's an open source image.
What constitutes Open Source or not is very clear now for software, but for other artifacts, that might not be so clear.". They may acknowledge open source beyond software, they don't consider working on it.

- 2026: 6 Month of wikipedia page reoriented toward open source resources, my research paper in front of the place. The silence again. Even had a meeting with the OSI director when I published my paper, not really engaging with the reasoning just saying management of source files with other resources is different (waiting more details on it).
[https://lnkd.in/eJQfjcFH]

If this organisation was really looking to understand open source, they probably wouldn't have missed all of this.

The open source community follow a blind authority with convictions ❤️

Erratum: Some coma of the OSD changed to accomodate some licences

As #Route66 marks its 100th anniversary, the Kansas City Fed’s Byways Report explores how tourism can serve as an economic development tool for rural communities.

Discover how communities can turn the place and stories that make them unique into opportunities for local growth.

Take the scenic route to rural prosperity: https://bit.ly/4kGE6Qs

I applaud the courage in this commentary about the state of academic publishing in our field!!
AI is not the problem; we are, as put eloquently by Mike Schmierbach:

"The current state of academic publishing means that in order for an individual to reap personal rewards, they must act in a way that damages the field as a whole. By incentivizing number of papers above all else, we have created the necessary conditions for the downfall of academic publishing."

I invite other journal editors and academic leaders to join this conversation so we can fix it!

https://lnkd.in/eFyVdpAz | 56 comments on LinkedIn

Elsevier generated more than US$3.5 billion in revenue last year. Springer Nature and Wiley also made billions. Academic publishing has become a huge business. Yet, one of its most important parts, peer review, is done mostly by researchers who receive no payment/incentives for their time and effort.

Every year, the number of research papers and journal submissions increases. Naturally, the pressure on reviewers is also growing. Finding reviewers, getting their consent, and receiving reviews on time have become increasingly difficult. I submitted a manuscript to a Wiley journal last April, but it only went for peer review two weeks ago!

At the same time, publishers charge US$2,000–4,000 in APCs for open-access publication. But the reviewers who give their valuable time to assess these papers receive nothing, not even a US$50 discount or voucher.

The problems are becoming more obvious. Reviews take months. Original reviewers are often unavailable for the second round, and new reviewers sometimes suggest fresh comments without properly considering the responses to earlier ones. Some reviewers seem to rely on ChatGPT and send AI-generated comments without checking them properly. Others make too many comments, many of which have little to do with the research objectives, methods, or findings. Sometimes, these comments even take researchers away from the main focus of their work.

Overall, something is clearly wrong with the academic publishing system. Publishers' revenues are growing, APCs are increasing, and more papers are being published. But are we actually getting better peer reviews? Long delays, inconsistent feedback, and the growing burden on unpaid reviewers are making the whole process increasingly frustrating.

🔎 Et vous, quel est votre profil de veille ?

Chercheur, organisé, analyste ou collaboratif… chacun a sa manière de faire de la veille !
Chez KB Crawl, nous savons que les besoins et les pratiques sont différents. Découvrez les 4 profils qui illustrent votre façon de rechercher, d’organiser et d’exploiter l’information. 💡
💬 Et vous, lequel vous correspond le mieux ?

💡Envie d'en savoir plus sur notre accompagnement en intelligence économique ? Découvrez KB Intelligence : https://lnkd.in/eKDNrKtR

#KBCrawl #VeilleStrategique #Veille #Information #IntelligenceEconomique

The University of Chinese Academy of Social Sciences' newly established school of rare and endangered disciplines will serve as an academic platform for deciphering ancient scripts, collating rare documents and conducting comparative research across civilizations.

🛠️ Combien de veilleurs ont un jour griffonné sur un coin de nappe l'outil dont ils rêvaient, avant de le ranger dans un tiroir faute de développeur, de budget ou de temps ?

J'en suis. J'ai longtemps imaginé des tableaux de bord, des cartes, des façons de restituer une veille autrement que par une liste de liens, et la plupart sont restées à l'état d'esquisse. Pour bâtir, il fallait savoir manier la truelle, ou payer quelqu'un qui la manie et ce n'était pas mon cas.

Avec le 𝘷𝘪𝘣𝘦𝘤𝘰𝘥𝘪𝘯𝘨, ce verrou a sauté, à condition de savoir ce que l'on veut. 𝗟’𝗮𝘁𝗲𝗹𝗶𝗲𝗿, la plus récente rubrique d'Outils Froids, est né ainsi.

Son premier outil s'appelle 𝗘𝘅𝗽𝗹𝗼𝗿𝗲𝗿 𝗹𝗲𝘀 𝗖𝗵𝗿𝗼𝗻𝗶𝗾𝘂𝗲𝘀 𝗱𝗲 𝗹𝗮 𝗥𝗲𝗰𝗵𝗲𝗿𝗰𝗵𝗲. Comme vous le savez peut-être, je repère chaque mois des publications scientifiques sur le renseignement, l'IA, la guerre cognitive et la désinformation, l'intelligence économique, le knowledge management et l'OSINT ; celles qui me semblent les plus intéressantes sont présentées dans mes Chroniques de la Recherche, dans Veillemag, partenaire du projet. L'outil réunit l'ensemble, soit déjà 320 références, chacune reliée à la chronique qui l'a présentée. On peut l'interroger de cinq façons :

→ un moteur de recherche
→ une carte en trois dimensions
→ une visite guidée mois par mois
→ un pont entre deux sujets
→ un diagramme

J'en présenterai une par jour.

Certes, l'IA ne fait pas tout. Elle a écrit le code, mais elle n'a pas décidé que la visite ressemblerait aux petits textes qui accompagnent les œuvres dans un musée, ni qu'il fallait un outil pour relier deux sujets que personne ne pense à rapprocher. Ces choix viennent du métier, c'est-à-dire de la curiosité, du goût des associations improbables, du souci de la forme qui rend une information lisible pour celui qui la reçoit. Bref, des qualités que l'on attend d'un bon analyste, et que la technique bridait dès qu'il s'agissait de les traduire en outil.

La créativité du veilleur butait sur un plafond technique. Il est aujourd'hui bien plus haut, et je ne suis pas certain que nous ayons encore mesuré ce que cela change pour nos métiers.

Et vous, quel outil auriez-vous construit depuis longtemps si la technique ne vous en avait pas empêché ?

#veille #vibecoding #IAgenerative #intelligenceeconomique #creativite

Pour découvrir la dernière version d'Outils Froids, suivez le lien dans le premier commentaire.

Academia rewards producing knowledge, not understanding it. We used to publish meaningful findings. Now we publish pictures of ourselves.
Once, publishing meant contributing something important.
Then publications became academic currency. More papers. Less contribution. Publication counts became the measure of success.
In the age of self-admiration, we publish, announce that we’ve published, and admire ourselves publicly for having published.
Now AI is accelerating the same dysfunctional system. It will drive publication numbers to absurd levels. Do we seriously believe anyone is reading all these articles?
AI will transform education: our morning seminars, journal clubs, and continuing medical education. It will also change how we publish and what gets published. Conventional structures will change. Knowledge is contextual. Checking that a study was conducted correctly and understanding what its findings add to existing knowledge are two different things.
But what happens when AI tells an author that their finding is not new—and has already been confirmed many times?
There goes the selfie!| 13 commentaires sur LinkedIn

❓ In your opinion, is the h-index a meaningful indicator of academic productivity, or does it fail to capture the true value of scholarly contributions?
📜 In our latest article, published in the prestigious Journal of Information Science, we critically examine the h-index, one of the most widely used metrics for assessing academic performance and research impact worldwide.
🧮 Drawing on both quantitative and qualitative evidence, we explore the strengths and limitations of the h-index and evaluate alternative performance assessment models. By challenging conventional approaches and highlighting the multidimensional nature of academic achievement, our study offers a fresh perspective on how scholarly productivity and research impact should be assessed.
🙏 I would like to express my sincere gratitude to Prof. Dr. Pinar Mert Cuce for her invaluable guidance, insightful contributions, and extensive expertise throughout this research.
https://lnkd.in/eXeDte5X
Have a look, and let the fake heroes crack! | 20 comments on LinkedIn

📊 𝗩𝗼𝗶𝗿 𝗲́𝘃𝗼𝗹𝘂𝗲𝗿 𝗹𝗲𝘀 𝘀𝘂𝗷𝗲𝘁𝘀 𝗱’𝘂𝗻𝗲 𝘃𝗲𝗶𝗹𝗹𝗲 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗾𝘂𝗲

Dans les années 1980, cartographier la dynamique d’un champ de recherche demandait un laboratoire et un logiciel dédié. C’est la méthode des « mots associés » de Michel Callon et Jean-Pierre Courtial, que j’ai appliquée à la veille scientifique que je mène pour Veillemag.

𝗟𝗲 𝗱𝗶𝗮𝗴𝗿𝗮𝗺𝗺𝗲, cinquième outil de L’atelier, range les familles de thèmes en quatre cases :
→ les sujets centraux et bien structurés
→ les niches pointues
→ les carrefours encore peu structurés
→ les sujets en marge

Juste en dessous, un diagramme de Sankey (que l'on devrait plutôt nommer diagramme de Minard, mais c'est une autre histoire) suit ces familles mois après mois. On les voit grossir, fusionner, se diviser ou s’éteindre. Un clic sur l’une d’elles montre d’où elle vient et ce qu’elle devient.

Comme les autres outils présentés cette semaine, le diagramme décrit ma veille, c’est-à-dire mes choix de lecture, et non l’état de la recherche. Si une famille s’amincit, les chercheurs s’en détournent peut-être... ou j’ai simplement regardé ailleurs ce trimestre-là.

Et vous , quelle méthode, découverte un jour dans un manuel et jugée hors de portée, aimeriez-vous enfin mettre en pratique grâce à l'IA ?

Lien vers le diagramme en commentaire ⬇️

#veille #scientometrie #intelligenceeconomique #vibecoding #IAgenerative

It is exciting seeing interest in #serendipity by researchers from so many disciplines and that #LIS and my work in information encountering are contributing to these efforts from the information behavior perspective.

Researchers conducted the first large-scale study of how often clinicians use artificial intelligence tools in patient care, who those clinicians are, and what they are using it for.

The study, led by researchers at Harvard Medical School and Brigham and Women's Hospital, looked at how 950,000 physicians, physician assistants, and nurse practitioners in the United States used OpenEvidence, a popular AI-enabled clinical information platform, over the last two and a half years.

The data showed that queries were frequently for aiding in treatment decisions rather than diagnostic tasks and that questions were often about subjects that were outside of the practitioner’s specialty.

“AI tools have become integrated into routine in clinical practice faster than almost any technology we’ve seen, but until now we’ve had very little data on how clinicians actually use them,” said first author Vishal Patel, MD, MPH, HMS clinical fellow in surgery at Brigham and Women’s. “This research gives us an early picture of where these tools fit into clinical decision-making and where we need to study their accuracy and effects on patient care next.”

The next step will be to examine whether use of these tools improves clinical care, the researchers said. | 13 comments on LinkedIn

Evidence Hierarchies are such a lightening rod in the world of evaluation. i’m not re-posting this one to trigger old debates but to point out a few features of this one.

first, the main purpose of this hierarchy is to help avoid over-claiming in conclusions. i agree with the post that it is all too easy to undo careful explanation in the methods and results with more sweeping statements in the conclusion (and executive summary).

second, i appreciated the cautions at the end of the post. that is, this hierarchy is only relevant for questions of causal attribution. if you want to answer other questions, then look to other approaches.

and quality still matters - a well-designed cohort study can tell you more than a poorly designed RCT. randomisation is powerful but we shouldn’t fetishise it!

finally, it’s useful to be reminded of the suite of go-to methods (of causal inference) for different fields. there’s no mention here of discontinuity design, diff-in-diff, synthetic controls or interrupted time series, which are common in the social sciences. not a criticism, just a reflection on differences between fields.