**Update** the app was down for a bit, there was more traffic than expected and the server couldn't handle it. I've upgraded to a larger machine. Let me know if you encounter any issue!
I'm excited to share mypapers.ai, though not quite in the way I originally envisioned. This project began as an ambitious AI-powered research assistant but evolved into a humbling learning experience about product development, over-engineering, and the importance of solving real problems.
What mypapers.ai currently does:
Visualizes daily arXiv papers as a graph with their references
Offers a simple interface to explore recent publications
What I learned along the way:
The dangers of solution-first thinking
The importance of "dogfooding" your own product
How over-engineering can derail a project
The value of starting simple with familiar tools
I've been building this project in public, sharing both successes and failures. While the outcome isn't what I initially planned, the journey has been incredibly educational.
I invite you to try mypapers.ai and share your thoughts. I think the most valuable this product has is the blog post I wrote about the journey. You can read it here:
https://medium.com/@pol.avec/bui...
@pol@pol_avec Congratulations on the launch! As a fan of the "Two minute papers" youtube channel, I approve this project! What would be also cool is to have a some sort of summary of the most influential papers, e.g. weekly, that you can display on the right hand side of the site (where is empty initially).
The summary you can automate with LLM i guess, or just write yourself :)
Btw, the site is currently not loading for me, but I understand you have some backend problems - hope you fix this soon!
This is an interesting tool, especially for those of us following LLM developments closely. How effective has the visualization been in helping users identify significant papers and their connections? Any plans to include more interactive features in the future? Congrats on the launch!
@kyrylosilin The most useful for me has been to see which are the most referenced papers day by day. That helps me prioritize which are the basics I should read to keep up, instead of trying to read each day's papers.
Regarding the features, so far I've been the only one using it so I need feedback on which features are needed. Some ideas in the pipeline are:
* allow users to expand the graph by clicking on nodes
* allow search / retrieval using LLMs (which is already in the back-end, not front-end)
* visualize connections by author
Is there any feature you'd find useful? what are your thoughts?
Congrats on launching mypapers.ai - seems like a handy tool for LLM research, making it easier to track and visualize new arXiv papers, right? And you also have a great and honest story in that medium article!
Say, how does mypapers.ai handle papers with numerous references? I'm interested to see how this simplifies workflow. Great job!
@adamyork not sure what's your question but here's how it works. Each day I add all published papers of that day. For each paper on that date, then I also add all the papers that the paper cites. That way you start with recent papers, and organically build the the references. Doing this you won't have old non-cited papers, but I thought it was a fair compromise given that the solution is to explore a network (non-cited papers are not connected in that case)
the app was down for a bit, there was more traffic than expected and the server couldn't handle it. I've upgraded to a larger machine. Let me know if you encounter any issue!
Report
Congrats on launch mypapers.ai! The idea of visualizing research papers is fantastic.
Fewsats
notclass
Fewsats
Fewsats
AI VisionBoard
Telebugs
Fewsats
Telebugs
Fewsats
Fewsats
Fewsats
Fewsats
Fewsats