Justus Mattern @ ICML (@MatternJustus) / X

Justus Mattern @ ICML

1,173 posts

Justus Mattern @ ICML

@MatternJustus

Co-Founder @ProximalHQ | prev. research @PrimeIntellect, @MPI_IS and built revideo

San Francisco, CA

justusmattern.com

Born November 27, 2002

Joined March 2021

Pinned
Justus Mattern @ ICML
@MatternJustus
Apr 16
Introducing FrontierSWE, an ultra-long horizon coding benchmark. We test agents on some of the hardest technical tasks like optimizing a video rendering library or training a model to predict the quantum properties of molecules. Despite having 20 hours, they rarely succeed
272K
Justus Mattern @ ICML
@MatternJustus
Jan 4, 2025
The only piece of advice I give to undergrads that want to get into research is to cold email PhD students with a good track record. Most undergrads are bottlenecked by research ideas whereas good PhD students have way too many ideas that they cannot execute. If you can code
438K
Justus Mattern @ ICML
@MatternJustus
Jul 21, 2025
How to train the best non-reasoning model: 1. Gather a reasoning dataset 2. remove <think> tokens 3. Train the model
87K
Justus Mattern @ ICML
@MatternJustus
May 17, 2025
is this a viable path to making $10M: 1) team up with ai friends 2) raise seed 3) build RL envs for white collar tasks (read xlsx files, CRM data entry, etc.) 4) fork verl, train models 5) outperform o3 on said tasks 6) get acquired to join salesforce ai research
84K
Justus Mattern @ ICML
@MatternJustus
Feb 3, 2025
O1 Visa approved. It's time to build 🇺🇸
56K
Justus Mattern @ ICML
@MatternJustus
May 20, 2025
me at the office, grinding ablation experiments the unemployed friend on a tuesday night:
60K
Justus Mattern @ ICML
@MatternJustus
Jul 5, 2025
Highest leverage thing unskilled engineers can do rn to contribute to frontier AI research is vibecoding RL environments
80K
Justus Mattern @ ICML
@MatternJustus
Aug 7, 2025
The biggest lever towards ASI is right in front of you (RL environments) but you feel too smart to work on data
53K
Justus Mattern @ ICML
@MatternJustus
Oct 28, 2025
Update: After an incredible year at @PrimeIntellect, I have decided to take my next step in August. Grateful that I got to work with such a talented team and build the best open-source RL infra! For now, I'm continuing to work on RL for coding agents. Will share updates :)
132K
Justus Mattern @ ICML
@MatternJustus
May 12, 2025
We went from our first line of code for prime-rl to releasing INTELLECT-2 in around two months. Now that our infra is in place and proven to work, I’m very optimistic that it’s only a matter of time until we will catch up with frontier labs. Some thoughts on this release ⬇️
Prime Intellect
@PrimeIntellect
May 12, 2025
Releasing INTELLECT-2: We’re open-sourcing the first 32B parameter model trained via globally distributed reinforcement learning: • Detailed Technical Report • INTELLECT-2 model checkpoint primeintellect.ai/blog/intellect…
113K
Justus Mattern @ ICML
@MatternJustus
Oct 17, 2025
my turn: if you are interested in working on coding agent research and being a core contributor to what will be an impactful paper, some Stanford friends are working on an industry collab and are looking for motivated researchers to join! Can be paid internship if full-time, DM
Justus Mattern @ ICML
@MatternJustus
Jan 4, 2025
The only piece of advice I give to undergrads that want to get into research is to cold email PhD students with a good track record. Most undergrads are bottlenecked by research ideas whereas good PhD students have way too many ideas that they cannot execute. If you can code
60K
Justus Mattern @ ICML
@MatternJustus
Jul 17, 2025
RL with predefined tools does not matter in the long term, the most bitter lesson pilled approach is giving the model a single universal tool (a computer)
OpenAI
@OpenAI
Jul 17, 2025
Replying to @OpenAI
ChatGPT agent’s capabilities are reflected in its state-of-the-art performance on academic and real-world task evaluations, like data modeling, spreadsheet editing, and investment banking.
40K
Justus Mattern @ ICML
@MatternJustus
Jan 4, 2025
Replying to @MatternJustus
Implementing experiments that a competent PhD student gives you is a great way to get your first papers. As you start reading papers on your own, intuition for your own ideas will come automatically
24K
Justus Mattern @ ICML
@MatternJustus
Jan 24, 2025
Time to share this again: DM me if you want to join our reasoning team. If we succeed, your name will be on the paper of a fully open (code & data & weights) reproduction of o3 trained across globally distributed compute by the end of the year. Here's why I think it'll work:
Prime Intellect
@PrimeIntellect
Jan 22, 2025
Replying to @PrimeIntellect
Join Us: This year, we will train a fully open reproduction of o3, enabling everyone globally to contribute compute resources. If you want to be part of this, DM us for collabs @johannes_hage @MatternJustus or apply here: jobs.ashbyhq.com/PrimeIntellect
48K