Since 2019 I maintained instructions and ROOT build that runs on Google Colab. But every couple of months I had to re-do it because Colab updates. But Now I automated this process and it will always be updated and published here melashri.net/ROOT. I wish it could be an official page on root.cern.

My dream app which I want to write someday will be an arxiv and CDS/inspirehep reader that you subscribe to categories and topics(keywords) and it will give you latest papers per day organized for you just to read. Without emails or LLM summary or all of that.

I have a side project. It is custom semantic search for LHCb papers. The end goal is to construct a knowledge base for LHCb (not only papers). Currently I have a web app that takes a query and then give you the 10 papers relate most to your query. Looking for a small number of beta testers.

First time to feel privileged in my life was yesterday during my French visa appointment. Once they knew I have CERN affiliation. I was prepared with every official paper I had on me and they just asked for passport and invitation letter. Other people before me had to submit dozens of papers.

My theory is that overleaf issues is related to their general deployment of AI error assistant. I was part of beta test and now got an email saying they are moving it from lab to general availability. Also my experience with it is that it is not worth it. And it is $60 annual add-on.

Overleaf is down. If you have a paper deadline then my condolences. This is a good chance to remind you to use sync to Git/Github or Dropbox for all your documents on Overleaf. This could come handy in situations like that.

I hate to have more than 10 open tabs on my browser. But during my shifts (which I do online) I have to open many more than that. I get confused all the time. My solution is is use zen browser "Firefox fork" and then assign workspace for shift related stuff and have a quick look every hour.

So Deepseek announced their R1 lite model which use reasoning and is competing with o1 performance. Qwen2.5 line of models are impressive specially the coder versions. There are rumors about qwen3.0 coming. Open source (weights) models seems to be living the golden age. If only GPU prices decrease 🤦

Ever wondered about the scale of cyber threats faced by major research institutions? Let's delve into some estimations regarding CERN's daily cyber challenges.

This is a thread about estimating how many cyberattacks CERN gets per day. This will be a back-of-the-envelope estimation.

1/8

y than mine on why #CERN experiments will prevent trigger development computing nodes from connecting to outside world?

My theory is to prevent abuse in the form of mining cryptocurrency to help pay the rent for poor PhD students /s.

Guess what just arrived via mail today. It is our beloved PDG

I put together a starter pack for people working in particle physics experiments (for now only collider based). Suggestions are very welcome! ⚛️ go.bsky.app/TFne5pM

I'm working on a Starter Pack to help identify the people from CERN community so that it is easier to form network of people interested in CERN and form hopefully a community here on Bluesky. Feel free to suggest names.
.

go.bsky.app/8kj2G8A

Working with GPUs in both personal projects and at LHCb’s GPU-based Trigger has shown me the impact of real-time, high-throughput data processing. In high-energy physics, we need fast, efficient filters to capture meaningful events from massive data streams without overloading resources.

Introduction: I'm a PhD candiate at UC, I'm working as part of @lhcb.bsky.social studying rare B decay. I also work on LHCb trigger and neural networks. I like contributing to open source projects, Homelabs, Selfhosting and Hardware. I spend most of my time working with CUDA and C++.