Julien Vanegue
@jvanegue
Followers
6K
Following
1K
Media
328
Statuses
9K
CTO Office / Head of Infra & Security Research @Bloomberg. Interested in mathematical techniques for software, systems, and network analysis at world scale.
New York City, NY, USA
Joined April 2010
I am hiring a full time researcher for our team in Bloomberg. Candidates must be university professors or experienced industry researchers in the following fields: Network design & verification, AI for infrastructure & security, Compute & Storage, Systems & PL. Reach out!
3
16
40
I keep coming back to this thinking I love this prioritization criteria. What is yours?
Here is one simple trick I use that works for both humans and AIs: When comes time to prioritize tasks for my own backlog, I follow one simple rule: always prioritize tasks that will unblock others. This is the only way to scale.
0
0
0
If you are a researcher or technical leader recently impacted by layoffs at Meta or Amazon, know that we are hiring at Bloomberg in NYC. Reach out to me and I will help.
0
9
21
Interesting graph theory fact: reachability is transitive, but unreachability is not. R(x,y)^R(y,z) => R(x,z) But U(x,y) ^ U(y,z) does not imply U(x,z) This is quite unfortunate for network security. Has anyone documented this? I cant possibly be the first one to notice this.
1
0
6
The work of Tavis Ormandy has been a strong influence on my personal research program. Lately, our work on automatically finding divergent execution bugs was directly inspired by one of Tavis’ infinite loop vulnerability in openssl’s elliptic curve implementation. Hat tip 🎩
A personal update... after nearly 20 years at Google, today is my last day! I'm going to be working on independent research for the foreseeable future, then who knows! I've worked with so many talented people, made so many friends and seen incredible research over the years 🫡
0
5
49
The abstraction-refinement dataset is now complete for interpolation (passes training examples) of @arcprize 2 evaluation puzzles. Now it's possible to begin analysing the data to spot patterns and generate informative reports: Statistical Summary: From my analysis of 111
1
3
4
My brain broke when I read this paper. A tiny 7 Million parameter model just beat DeepSeek-R1, Gemini 2.5 pro, and o3-mini at reasoning on both ARG-AGI 1 and ARC-AGI 2. It's called Tiny Recursive Model (TRM) from Samsung. How can a model 10,000x smaller be smarter? Here's how
348
2K
12K
MIT open courseware is a treasure waiting to be (re)discovered. Such a treat to get free access to @MIT's STEM material:
ocw.mit.edu
MIT OpenCourseWare is a web based publication of virtually all MIT course content. OCW is open and available to the world and is a permanent MIT activity
0
3
12
We were walking on NYC’s Bleecker street when I realized I’m a damn lucky guy to hang out with Shubham Ugare (@TechAtBloomberg intern 2024, now at @Meta) and Tarun Suresh (intern 2025, now back at UIUC). “Let’s immortalize this!” I asked.
0
0
5
Interested in practical tips for using agentic LLMs for vulnerability detection? Check out this tech report from the Fuzzing Brain team, who reached the finals of AIxCC. It was a pleasure to be involved in this effort! https://t.co/KLmpntJoEN
0
18
79
I am happy to report that my infinite loop checker has reached mainstream @fbinfer — Huge thanks to David Pichardie for merging the PR.
github.com
Pulse-infinite, the new INFINITE_LOOP checker in Pulse Tool Behind paper: "Non-Termination Proving: 100M LoC and Beyond" https://www.arxiv.org/abs/2509.05293 Please see CONTRIBUTI...
0
1
7
I just wrote a simple and dumb program. Do you think it will terminate for any value of x? f(int x) { while (x != 1) { if (x % 2) x = x * 3 + 1; else x = x / 2; } }
5
0
3
The numbers of @NeurIPSConf accepted papers about code AI is ridiculously small (similar to last year's). What would you say got the biggest part of the cake? https://t.co/S7SdWtF4Ib
0
0
1
Submit your early research to the 3rd POPL workshop on the theory and practice of static analysis (TPSA) : https://t.co/v7g8i5sCHX - Call for presentation deadline: Oct 24th 2025
0
0
0
Join us in-person at @Bloomberg's #Quant (BBQ) Seminar on Monday, Sept. 29, 2025 @ 5 PM EDT, which will feature a keynote about #QuantumComputing by @Dr_Chetan_Nayak of @Microsoft & @UCSBPhysics + lightning talks https://t.co/gOOOLsTGTM
#QuantFinance #QuantitativeFinance #Quants
0
1
1
Here is one simple trick I use that works for both humans and AIs: When comes time to prioritize tasks for my own backlog, I follow one simple rule: always prioritize tasks that will unblock others. This is the only way to scale.
0
0
5
It is impressive to meet one of these native AI-era programmers. You’ll recognize one when you see one! These are not “vibe coders” but capable programmers compressing time to market by 5-10x while coding with chatgpt. Most AI-era coders I met are under 30.
0
0
3
I have doubt on the narrative seen in so many recent news articles saying AI is replacing junior engineers in favor of more senior engineers. The best AI-era programmers I have met are those born with it: new grads. Many senior engs still mistakenly doubt AI or under-utilize it.
1
0
6
Testing modular abstraction ideas to decompose a real complex @arcprize puzzle into more manageable subproblems. Lots to learn even from a single task. First tried with ideas from vision, then some generic pattern recognition simply on numbers and coordinates. On the fence on
3
2
7
"How I became a story teller?" by @SumitGulwani at @MSFTResearch is one of the most useful talk I ever watched. This will for sure lead me to improving my public speaking skills. Thank you Sumit!
2
2
12