Murfy AI - Write, review, and publish to arXiv 10x faster

@shounan_an One of the devs behind Murfy here!! It’s been a lot of fun (and a lot of LaTeX debugging) getting to this point. Excited to finally share it with everyone. Let us know what’s broken or annoying!!

I work on Murfy's backend. My side of it comes down to one thing: your compile should be fast, reproducible, and completely unremarkable. If it's ever not, this thread is a good place to tell me.

@shounan_an Having real-time collaboration alongside contextual AI agent editing solves the biggest bottleneck in group papers. congrats🙌

@shounan_an the fact that it can fix compile errors is actually pretty interesting. does it just fix the error, or can it figure out when the fix has messed up the formatting somewhere else?

@shounan_an Congrats on the launch! The idea of having AI actually work through the research writing workflow rather than just acting as another writing assistant is really interesting.

The LaTeX compile error fixing and handling reviewer comments especially caught my attention. Wishing the team a great launch and excited to see how Murfy evolves!

@sabber_ahamed Thank you for the kind words. Hearing this from a long-time LaTeX user means a lot to us. We know there's still plenty to improve, and we'll keep working to make it better. Thanks for the good wishes.

@sabber_ahamed Thank you! It’s always great to hear from a fellow LaTeX user. Hope Murfy can be useful for some of your future work!

@sabber_ahamed Thank you so much Sabber! Wish you good luck too :)

@rick_segal1 Honestly, Overleaf is the standard for online LaTeX editor, and it has years of maturity and a huge ecosystem behind it.

The main difference is that Murfy is built around paper-writing process, not just editing LaTeX. Its AI works more like an agent than a chat box: it reads the project files, suggests changes as a diff, compiles the paper, and works through errors until they’re fixed. You can accept or reject each change.

Murfy can also review the finished paper from the compiled PDF. It checks whether the claims are supported by the paper’s own evidence, challenges weak assumptions, and returns a referee-style report with a rating and confidence score.

We also don’t limit collaborators on the free plan, so you can invite your whole lab.

Where Overleaf is still ahead is maturity and integrations. We’re much newer, and we know we still have plenty to improve.

@shounan_an For refrence

@raunaks_99 

Thanks for testing it twice and sharing such detailed feedback. This is genuinely helpful.

1. About the "Murfy Guide" first page:

It can be added from two separate places, so it's easy to remove one and miss the other. Please check your main .tex file for both of these:

- \input{murfy-guide.tex}
- \murfyguide

Remove both and recompile. If the page is still there, please use "Report bug" in the profile menu so we can take a closer look at your project.

2. About the Korean replies:

That one's on us. Some of our internal instructions are written in Korean, and occasionally that leaks into the response. We're working on a proper fix.

Sorry for the inconvenience, and thanks again for taking the time to report this.

@natalia_iankovych That's not a feature we offer. Murfy is built specifically for academic research writing, where the goal is the opposite: helping researchers stay in full control of their own voice while using AI as an assistant.

For research papers, we'd always recommend that approach anyway.

@rizzykicks Great question! It varies a lot by experience level.

Junior researchers save the most time on LaTeX itself via code assistance and compile error fixing. The learning curve disappears fast.

Senior researchers get the most value from deep AI paper reviews, which catch weaknesses, reproducibility issues, and improve research papers quality from AI review feedback before submission. Iterating on that loop is where the real time savings stack up.

Advisors told us the biggest win is visibility, seeing who changed what and when through revision history, plus AI-verified reference checking that used to be pure manual work.

Which stage are you at? Happy to dig into what Murfy can do for your specific workflow :)

@ashish_khandelwal11 Great question! Murfy is model-agnostic by design, we select the best AI model for each step of the research workflow, including open-source models where they perform best.

One principle we never compromise on: your data is never used for AI training, regardless of which model runs under the hood.

We're also working on letting researchers choose their preferred model directly.

Stay tuned!

@yelyzaveta_kibets There’s no template library behind it. it works directly with the files in your project, including any .cls or .sty files you’ve uploaded. So custom templates generally work the same way as standard ones. But we’ve tested it more heavily with standard document classes so far.

If you have a custom template that trips it up, please send it to us. Those are exactly the cases we want to test more.

@galdayan That's actually the exact situation that gave us the idea-even the third submission part. We've also seen people ask Murfy things like, "Convert this ACM paper to the Elsevier template" or "Make this fit the KCC format."

We're working on it now. The idea is that you choose the target venue, and Murfy handles the class, packages, structure, and bibliography style, then fixes any compile errors along the way. You'll still get a diff so you can review everything.

Bibliographies are the tricky bit, since changing the .bst file doesn't always cover everything. We'd rather show you what changed than silently rewrite your references.

Still in development, but coming soon!

@galdayan Resubmission after rejection, almost the same paper, different format: pure busywork with zero intellectual content. You nailed it.

That feature is coming in Murfy very soon.

Would love to test it against a real case when it's ready, what venue are you targeting next?