Marginalia · July 2026

how not to be reviewer 2

Everything I wish someone had handed me before my first review — how to structure one, how not to be a jerk, and how to keep up with the literature without losing your evenings.

Someone is going to ask you to peer review a paper.

Maybe they already have.

You said yes, closed the laptop, and immediately felt a specific kind of dread.

Not because you don’t know the science.

Because you don’t know what you’re supposed to do with it.

Every field has a version of “Reviewer 2” — the anonymous, faintly cruel voice that asks for one more year of experiments for no clear reason.

Here’s the uncomfortable part. Reviewer 2 isn’t a person. Reviewer 2 is a role, and someone just handed you the script.

This is my attempt to write down what I wish someone had handed me instead, especially now that I’m trying to review openly, with my name attached, through places like PREreview.

Turns out a lot of smart people have already worked this out. James Fraser’s lab even runs an actual course on it every year. I’m mostly just the messenger.

the shape of a good review

Fraser Lab’s guide to peer review breaks a review into three parts, and once you see it you can’t unsee it.

Summary. Three to five sentences. What were they trying to do, did it work, what’s missing, why does it matter to the field.

Major points. The things that actually change whether the conclusions hold up.

Minor points. Everything else — confusing figures, typos, a citation doing more work than it should.

The trick is in how you write the major points.

Don’t write: “It would be nice to include another technique.”

Write: “Your data don’t fully support this claim. Consider adding [specific control], or narrowing the claim to [reduced interpretation].”

One of those is a request. The other is a complaint wearing a request’s clothes.

Most of the time, the fix doesn’t need a new experiment at all.

It needs a sentence rewritten so it stops promising more than the data can back up.

the FAST rules, so you don’t have to invent your own

ASAPbio’s FAST principles exist for exactly this moment — the one where you’re staring at a blank comment box on a preprint, unsure what’s fair game.

Focused. Talk about the science. Not the journal it might land in. Not the authors.

Appropriate. Professional tone, always. Say something if someone else’s review isn’t.

Specific. Say what would change your mind, not just that you’re unconvinced.

Transparent. Sign it if you can. If you can’t, say why. Either way, own it.

Notice none of these say “be harsh” or “be nice.” They say be legible. That turns out to matter more.

don’t be an a$$hole (they said it twice, so I will too)

Buried in the syllabus for Fraser Lab’s peer review of AI-in-biology course is a line that made me laugh out loud: don’t be an a$$hole, stated twice, for emphasis.

The longer version:

Assume you’ve misunderstood before you assume the authors are wrong.

Imagine receiving your own review. Would you want to?

The paper doesn’t have to be the paper you would have written. That’s not a flaw.

Name your own biases before they show up uninvited in your comments.

I’ve started using “would I be comfortable saying this to their face, with my name on it” as the actual bar. Turns out that’s most of what open review asks of you anyway.

if you’re the one being reviewed

The companion guide to responding to reviews has one idea that rewired how I read criticism of my own work.

Clarify. Don’t defend.

If a reviewer misunderstood something, that’s rarely their failure. It’s usually a sign the manuscript wasn’t as clear as you thought it was.

Practical version: wait 24 hours before responding to anything. Put every review in one document. Color code — their comments in black, your response in blue, new manuscript text in red. Answer every single comment, including the nice ones.

Reread your response as if you were the reviewer. If it would annoy you, rewrite it.

reviewing out loud, on purpose

This is the part that used to make me anxious and now makes me a little proud.

James Fraser’s lab only reviews papers that are already preprints, and posts every one of their reviews publicly. Not because it’s easier. Because, as their broader philosophy puts it, the actual goal is to “get the work out there for others to build on, not to publish in journals.”

PREreview’s Open Reviewers program trains people to do exactly what I’ve described above, and to do it in public, with a name and a DOI attached, while actively watching for the biases that creep into anonymous review.

Nikolai Slavov’s argument in eLife is the case for why that discomfort is worth it: hiding review reports wastes real expertise, and legitimate problems raised in the open, on PubPeer, in comments, wherever, get ignored far too often when nobody’s obligated to respond to them.

Open review isn’t a nicer version of the old system. It’s closer to the actual point of review — making the record better, in front of the people who’ll use it.

If you’ve been putting off signing up for something like PREreview because open review feels higher-stakes than anonymous review, it is. That’s kind of the idea.

keeping up with the literature in about fifteen minutes

None of this matters if you never see the papers worth reviewing in the first place.

Fraser’s method for this is almost insultingly simple: an RSS reader, subscribed to your field’s journals and a few saved PubMed searches, read in three passes.

Scan 100% of titles.

Read the abstracts of the ones that survive, call it 20%.

Open the full text of those, another 20%.

Sit down and actually study maybe one or two papers a week.

Do the scanning on your phone while you’re walking somewhere. Do the abstracts while you’re waiting for coffee. Save full papers for when you’re actually at a desk.

He calls this “one of the most important and enjoyable responsibilities” of being a scientist, which is a generous way to describe doomscrolling PubMed. I’ve started to believe it anyway.

send this to whoever asked you to review something

If you’re anxious about your first open review, you’re not behind. You just haven’t seen the playbook yet.

None of this is mine. I just collected it in one place because I needed it in one place.

Consider this that place.

Sources, properly credited — this entire guide is a remix of other people’s thinking, not mine: