Claes Bäckman

It's that time of the year where new PhD students have to figure out how to do research. Luckily, the profession has put together a lot of resources to help us all become better researchers, and you can find my collected list of resources here: sites.google.com/view/claesback…

Submit full papers or extended abstracts here: safe-frankfurt.de/news-latest/ev…

📅 Submission deadline: 15 April 2026 📣 Notification: 30 April 2026 📍 Frankfurt School of Finance & Management 2 slots for PhD candidates with limited travel support.

📢 CfP: Housing Market Frictions and Access to Homeownership 1st Annual Workshop in Real Estate Finance · 24–25 Aug 2026 · Frankfurt Keynote speakers: • Timothy McQuade (UC Berkeley, NBER) • Andreas Fuster (Swiss Finance Institute at EPFL, CEPR)

@ChenhaoTan Thanks! I used it for my paper, and we also used Refine. I think the comments were pretty aligned. But I have not done a really formal comparison, I mostly wrote this for myself and thought I should share it

@ClaesBackman Just added! Great resources, did you compare this with Refine?

Peer review is facing a death spiral, and AI production tools are speeding it up. AI-assisted reviewing is necessary and should be open. We built OpenAIReview: open AI reviewing for everyone, for the cost of a coffee. openaireview.github.io/blog.html 🧵

@peternemes That is cool! Hope it works well!

@ClaesBackman Made an update to this Claude Skill, for use in condensed matter research. Specifically in scanning tunneling microscopy, spectroscopy, etc github.com/zrbyte/AI-rese…

I just published a free Claude Skill that generates feedback on your own academic papers. Link in the comments below.

PS. I should note that the impact of AI on research quality is often mentioned in the posts about publishing. People are clearly aware of this. On the margin, though, I thought there was space for a slightly more optimistic take.

Claude Code and similar tools are tremendously useful for generating feedback on research, and I’m optimistic that this will lead to better research. I wrote about this and what the democratization of research quality will mean for us as researchers.

There has been a great debate recently about the impact of AI on publishing, driven by great posts by @causalinf and @akoustov. Within this debate, I think the positive impact of AI on research quality should be emphasized a bit more.

I started building this for myself to get better AI feedback on political science research. After seeing @claesbackman’s econ-focused repo, I pushed it further into PoliSci Review: a journal-aware, LaTeX-first skill for political science drafts.

@Cemmd Looks very cool!

@JeremyNguyenPhD I can't take all the credit here! Claude did a lot of the formalizing of the task. I asked for a setup where each agent covered different tasks, and then I specified the task. Claude filled in the details. With some iteration, this is what came out. These tools are really amazing

@ClaesBackman PS. I'm so curious, Claes: was it that you've explained these steps many times to your PhD students, or did you explain them like this to your RAs? it's so methodical, and so many other people would just be unable to explain it, like a tacit type knowledge

Want a Claude Code skill to critiques your academic papers? Launches 6 parallel agents, each checking one aspect. Worth reading Claes's files here, just to see how clearly and methodically he decomposes it into such clear instructions. It's masterful. Thank you, Claes!

@MolBioMike Maybe I should have! But I like to think about this tool as a very detail-oriented and careful supervisor. It’s mostly trying to help get rid of stupid mistakes!

@ClaesBackman I hope you called it "Reviewer 2"

@cblatts I created something like this to generate feedback on different parts of research papers. Link here: github.com/claesbackman/A…

Finally, just like you created a Prompt Engineering project folder with instructions + preferences + best practices, you could create a Critique project folder. I haven't tried this myself (because I do it in Claude Code), but it should work.

So, you do NOT need Claude Code for better prompts or get "agents" (or personas) to iterate, critique & refine your ideas & investigations. Here are a few things I've done to get better feedback from Chatbots, whether it's for planning a family vacation or doing work research.

@sergii_x I didn't put that in, this is purely to provide feedback on existing materials. Citations and references should probably be a different skill!

@ClaesBackman By any chance, is it able to help with the citations?

You can also specify which journal you are targeting. You can install it in claude.ai in two minutes. Feedback welcome!

With the skill, you just type “/review-paper” and all agents will launch. The end result is a report with specific feedback for your paper. If you have a claude.md file, you can also add project-specific context that makes the feedback better.