Hi
I’m Rodrigo.
I’m building PaperStack because peer review is too important to be this slow, opaque, and weirdly dependent on unpaid goodwill.
The basic idea is simple: if you want your preprint reviewed, you also review someone else’s. That makes peer review faster by creating review supply instead of waiting for journals to beg busy people. It makes review more transparent because the output can become public, not hidden in private editorial inboxes. And it makes reviews better because PaperStack matches reviewers carefully and evaluates the reviews themselves, filtering out low-effort feedback and learning who gives useful criticism.
The ultimate goal is bigger than faster peer review. PaperStack aims to become a shared peer review layer, independent of publishers. Instead of each journal owning its own obscure and inefficient review system, preprints should be reviewed by the research community. After that, journals can choose what they want to publish, not what they want to accept for review and maybe, eventually, possibly publish after everyone has grown more gray hairs.
That is challenging, obviously. A system like this has to match expertise well, screen conflicts, create real incentives, and keep reviews useful without becoming another opaque gatekeeper. But thanks to preprints becoming normal, research communities already discussing work in public, and LLMs making better matching and quality checks possible, it finally feels possible.
If you want to talk about PaperStack, you can write to me at rodrigo@paperstack.pub.
Bests,
Rodrigo Rosas-Bertolini
How it works
1. Start with public preprints
PaperStack runs peer review in weekly rounds. Each Saturday, it builds a new pool of public preprints from preprint repositories and journal-club communities where researchers are already looking for feedback.
Authors can also ask PaperStack to include their preprint by writing to review@paperstack.pub. The goal is to start each round with work that is public, recent, and ready for serious review.
2. Group papers and authors by expertise
PaperStack analyzes each preprint to understand what kind of expertise would be needed to review it well. It then groups the preprint with similar work, so each review round has a map of which papers belong near each other. If a preprint sits between fields, PaperStack can treat it as needing more than one kind of reviewer expertise.
Then PaperStack looks at the researchers who authored the preprint. It uses their publication history and public metadata to understand what they work on and which kinds of papers they could review in return.
Authors are then invited to join a reciprocal review round. They have one week to accept the simple deal: review other papers, and your paper gets reviewed too.
3. Send each participant a shortlist
Once authors accept the invitation to join a review cohort, PaperStack uses network analysis to send each participant a shortlist of potential preprints to review. Participants can choose from that list, or decline if none of the options is a good fit.
The shortlist is a preference system, not a final assignment. A participant might receive ten possible preprints, rank their top four, and ultimately be assigned two depending on availability and the shape of the whole review network.
PaperStack also checks relationship signals, such as co-authorship, institutional overlap, and other connections, so the system does not become a closed circle of friends reviewing each other too generously.
4. Build reviewer chains in a cohort
Once researchers choose their preferred papers, PaperStack builds reviewer chains. In the smallest version, three people can each review one paper and receive one review without reviewing the person who reviews them.
Scaled carefully, this means most participants are reviewed by someone qualified and not carrying obvious favor or animosity. PaperStack then tells each researcher which paper(s) they have actually been assigned. Authors who cannot be fit into reviewer chains are informed and removed from the cohort, but they can opt in to be considered for the next one.
PaperStack also gives participants a deadline and asks them to submit their reviews by email to review@paperstack.pub. A review template can be suggested by the author, but it is not enforced.
5. Evaluate and filter reviews
PaperStack checks whether each review is useful, fair, specific, and written by someone actually engaging with the work. Rude comments, lazy summaries, and empty AI-generated reviews are rejected.
LLM assistance is welcome when it sharpens a reviewer's own judgment. Outsourcing the review to a model and sending back empty slop is not. If a review fails that quality check, PaperStack rejects it and informs both the author of the review and the intended recipient.
If a participant submitted a valid review but did not receive one back because the review intended for them was not delivered or did not pass quality checks, PaperStack gives them a credit: an IOU for a future cohort. This means that, in a future cohort, they will receive an additional review without having to provide one.
The authors of reviews that fail quality checks, and participants who miss review deadlines, are banned from using PaperStack in future cohorts.
6. Close the review loop
PaperStack emails the accepted reviews to the authors of the preprint and asks them to give feedback on the reviews themselves.
The same email asks whether the authors want the reviews they received to become public, and whether they want to continue the peer review process with the reviewers who reviewed them.
If the reviewers accept, PaperStack carries that preprint-reviewer link into a new cohort as part of the reviewer chain. This lets authors receive multiple rounds of peer review from researchers who have already shown that they can give useful, actionable feedback.
7. Post the approved reviews
When authors approve reviews for public release, PaperStack posts those reviews where the authors want them to appear, when that destination is supported.
That could be a preprint server, a journal club platform, or another place where the research community already discusses the work. Not every platform is supported yet, but PaperStack will do its best to build broad compatibility.
PaperStack posts reviews on behalf of reviewers to control formatting, make sure approved reviews are not altered, and maintain reviewer anonymity when reviewers ask for it.
Roadmap
1. Refine the review process
PaperStack currently has the system that does the core review loop: collecting preprints, matching expertise, building reviewer chains, filtering reviews, and letting authors decide what happens to the reviews they received. The next work is to improve and refine that system. PaperStack needs to get very good at managing the peer review process and making sure the guardrails work. Growing the community is a necessary goal for achieving the following stages of the roadmap.
2. Learn who reviews well
Once review rounds are reliable and frequent, PaperStack can start learning from the reviews themselves. It can compare reviewer feedback, incorporate author responses, and identify which reviewers consistently give fair, specific, actionable criticism. Over time, that creates a reviewer-quality layer: better future matching, stronger incentives for serious reviewing, and a reputation system based on the usefulness of someone’s reviews rather than just their publication record.
3. Journals come to the papers
Once PaperStack is processing a meaningful share of preprints in a field, journals can enter the loop. Instead of authors sending a paper to one journal at a time, journals could see community-reviewed preprints and make offers. Authors could choose between credible publication paths. That is the larger paradigm shift in scholarly communication: journals come to the papers, not the other way around.
Pricing
PaperStack is free for now while the reviewer network is still growing.
Once enough people are accepting review invitations and delivering useful reviews, participation in a review round will cost $9.
Right now, too few people accept and complete reviews, so the risk of not delivering is not consistent with the quality PaperStack wants to offer. Until that changes, I will run the service at my own expense.