The challenge for cross-disciplinary working groups is finding the right level of communication: too abstract and nothing interesting emerges; too technical and others can’t follow, via @mrillig.bsky.social
matthiasrillig.substack.com/p/how-to-wor...
How do you work with researchers from different disciplines?
Researchers from different fields can generate exciting new questions when they gather around a shared theme. 1/2
Writing a grant proposal?
✍️ Tip: Group similar tasks (like emails and administrative forms) and block out specific calendar time for them to stay efficient.
paperpile.com/g/how-to-wri...
What makes English written by a human sound human?
Lack of variation is one of the giveaways of AI-generated language, via @us.theconversation.com
theconversation.com/how-ai-engli...
💡 Quick tip: Save citations directly from a Google Scholar Labs conversation to your Paperpile library—with the AI-generated summary automatically added as a note.
An analysis of more than 500 scientific talks finds humor isn’t often used in them—and when it is, the jokes usually fall flat or elicit only polite chuckles, via @ScienceMagazine
www.science.org/content/arti...
Researchers are increasingly using LLMs to help format bibliographies for manuscripts. But these models sometimes generate non-existent academic references.
Tens of thousands of publications from 2025 might include invalid references generated by AI, via @nature.com
www.nature.com/articles/d41...
Thank you for recommending us, Claire!
Strict reproducible pipelines that always run from raw data can slow down analysis and break over time. 📊
A practical approach is to save the final processed dataset before plotting, which makes results easier to access in the future, via @clauswilke.com
blog.genesmindsmachines.com/p/creating-r...
Filler transitions, em dashes on every line, self-answered rhetorical questions...all classic tells of AI writing.
A Claude skill can remove these patterns, which is useful for paperwork like budget justifications you don’t want to write anyway, via @stephenturner.us
blog.stephenturner.us/p/deslop
🧮 Check things that should add up
⚠️ Never ignore code warnings
📊 Track datapoints like the number of rows
🔍 Investigate outliers
🧠 If something doesn’t make sense, keep digging
arjunrajlab.substack.com/p/transition...
AI is pushing computational biology into analytic abundance.
The bottleneck is no longer generating results: It’s deciding whether to trust them. 📊
5 habits for sense-checking outputs, via @arjunraj.bsky.social 1/2
AI is making it easier to analyze and stress-test scientific work at scale, which could improve rigor and reproducibility.
But there’s a risk: more data and better predictions don’t necessarily lead to paradigm shifts, via @asimovpress.bsky.social
Want early access? Contact support@paperpile.com to join the beta.
Learn more → paperpile.com/h/ask-ai/
Coming soon: Paperpile’s first AI-native integration ✨️
Send papers from your library directly to NotebookLM, ChatGPT, or your preferred AI assistant.
2/2 But it also risks pushing science toward “checklist” thinking, where easily measurable robustness is prioritized over judgment, creativity, and meaningful questions, via @jessicahullman.bsky.social
jessicahullman.substack.com/p/living-the...
AI is making it easy to run large-scale robustness checks like multiverse analyses on research papers, potentially transforming peer review and scientific scrutiny.
This could make it harder for fragile results to be published. 1/2
An analysis of submissions to bioRxiv shows steady growth in the number of scientists uploading their work. In 2025, more than 4,000 papers are posted each month, alongside millions of views and downloads, via @nature.com
www.nature.com/articles/d41...
AI can speed up parts of research, but when data are messy and fragmented, careful human verification is still essential, via @adamjkucharski.bsky.social
kucharski.substack.com/p/how-much-t...
Compiling research datasets from scattered reports can take months of manual work.
When the same task was attempted with an AI agent, it produced results quickly, but missed many cases and data sources. 1/2
A sabbatical is often seen as a time to produce—finish papers, push projects forward.
But it can also be a chance to step back, reframe your goals, and rethink how you want to work, via the Dynamic Ecology blog #AcademicSky
Open AI-assisted review tools could help researchers get better feedback, improve paper quality, and make the review process more scalable.
But final publication decisions should remain under human oversight, via @chenhaotan.bsky.social
openaireview.github.io/blog.html
Peer review is under strain.
Submissions are rising faster than the reviewer pool can handle. 1/2
Want to stand out in a PhD application?
Clearly explain your research direction, show how you think scientifically, and engage thoughtfully during interviews, via @nature.com
The binding constraint in science is shifting from production to evaluation.
The challenge is no longer generating papers, but deciding which ones matter, via @causalinf.bsky.social
causalinf.substack.com/p/claude-cod...
A researcher reflects on the moment their university email was set to be deactivated after a temporary contract ended.
After years moving between short-term positions, the loss of that address felt like losing a piece of professional identity, via @science.org
www.science.org/content/arti...
🔎 Pick one task that feels consistently draining
📝 Document the steps the next time you perform that task
🔁 Refine it over time
www.publishnotperish.net/p/the-hidden...
Routine research and teaching tasks become high-friction when you have to remember the steps each time.
Writing down the steps for repeatable tasks saves energy and keeps you from solving the same problems every semester, so you can focus on the work that matters, via @jennmcclearen.bsky.social
✍️ Academic writing tip: Your introduction isn’t a summary of everything ever published.
Its job is to clarify the conversation, identify the gap, and make your research question feel necessary. That requires interpretation, not a catalog of studies.
