Pinned Tweet
More Research Scientist positions opening up in Montreal!
We're looking for candidates with experience in video gen, RL, world models, and adjacent topics. Come join our World Modelling team at our Petite Italie / MILA office.
metacareers.com/profile/job_de…
Feel free to reach out!
We're looking for candidates with experience in video gen, RL, world models, and adjacent topics. Come join our World Modelling team at our Petite Italie / MILA office.
metacareers.com/profile/job_de…
Feel free to reach out!
3
16
181
15,052
Life update: a few months ago we made a Canadian. Today I became a Canadian myself. Proud to join this great nation that has treated us so well. Here’s to the future!
163
63
1,347
28,146
Daniel Severo retweeted
We've posted a 156-page tech report on Beam! We have good systems and it's worth reading. It covers all phases through pre-training, mid-training, SFT, RL, and OPD and goes into the infra, data, and scientific decisions we made throughout the process
📚 reflection.ai/documents/beam…
📚 reflection.ai/documents/beam…
We've released the weights for Beam!!
🤗 huggingface.co/reflection
📄 Model card: reflection.ai/beam-model-card
🧪 Try Beam in our playground & get an API key: platform.reflection.ai/playground
📚 AND we take open model safety seriously and have posted our safety report: reflection.ai/beam-safety-re…
19
83
588
43,258
Daniel Severo retweeted
You can now build a robotic scaling law with a cheap robot on 1 GPU. The recipe:
1. Collect 30mins of demonstrations with SO-101
2. Run inference with public RoboJEPA models (a few hours on 1 GPU).
3. Fit the scaling law.
details: robojepa.github.io
1. Collect 30mins of demonstrations with SO-101
2. Run inference with public RoboJEPA models (a few hours on 1 GPU).
3. Fit the scaling law.
details: robojepa.github.io
Introducing RoboJEPA, the first scaling law for multiembodiment world models.
We scaled JEPA predictors up to 8B params, showed scaling laws extrapolate and correlated compute scaling with planning performance.
We release all checkpoints and the code.
robojepa.github.io
5
7
73
3,881
Daniel Severo retweeted
Canada’s sovereignty depends on our ability to build, adopt, and govern AI on our own terms.
That’s why Canada is taking an equity stake in Cohere — one of the world’s leading builders of secure AI.
Sovereign AI. Built here at home.
Read more: pm.gc.ca/en/news/news-r…
That’s why Canada is taking an equity stake in Cohere — one of the world’s leading builders of secure AI.
Sovereign AI. Built here at home.
Read more: pm.gc.ca/en/news/news-r…
603
267
1,521
104,535
Daniel Severo retweeted
Congratulations, President @LulaOficial, on your re-election in Brazil.
The relationship between Canada and Brazil has flourished under our partnership. Let’s seize this opportunity to make it even stronger, together — to create major new trade opportunities, protect the environment, and build greater prosperity for both our peoples.
The relationship between Canada and Brazil has flourished under our partnership. Let’s seize this opportunity to make it even stronger, together — to create major new trade opportunities, protect the environment, and build greater prosperity for both our peoples.
2,279
8,068
56,082
1,054,874
Daniel Severo retweeted
Muse Spark Open -- weights releasing today -- runs great on new Mac Studios thanks to work with @johnternus and Apple.
1,952
1,745
32,087
4,901,159
Daniel Severo retweeted
Introducing RoboJEPA, the first scaling law for multiembodiment world models.
We scaled JEPA predictors up to 8B params, showed scaling laws extrapolate and correlated compute scaling with planning performance.
We release all checkpoints and the code.
robojepa.github.io
We scaled JEPA predictors up to 8B params, showed scaling laws extrapolate and correlated compute scaling with planning performance.
We release all checkpoints and the code.
robojepa.github.io
38
139
808
57,182
Daniel Severo retweeted
Losses were dominated by public sector jobs (particularly in education and health care) and self-employment.
Private sector jobs however, rose by 24,100.
📊
Private sector jobs however, rose by 24,100.
📊
#BREAKING: Statistics Canada says the economy shed 68,000 jobs in September, marking a second straight month of falling employment levels. bnnbloomberg.ca/business/econo…
34
37
326
71,256
Daniel Severo retweeted
Meta presents RoboJEPA
Scales JEPA predictors from 22M to 8B parameters (largest model trained to date), using data across 12 robotic embodiments, and studies how latent prediction error can forecast downstream planning performance
arxiv.org/abs/2610.10515
Scales JEPA predictors from 22M to 8B parameters (largest model trained to date), using data across 12 robotic embodiments, and studies how latent prediction error can forecast downstream planning performance
arxiv.org/abs/2610.10515
8
52
447
23,961
Gotta admit this is great. Goverment admin work really needs a rehaul. AI could make things so much more efficient and reduce the cost of goverment.
Can we get this too @MarkJCarney, maybe @cohere can help :D
Can we get this too @MarkJCarney, maybe @cohere can help :D
SECRETARY RUBIO on AMERICA.GOV: This is really going to be a revolution in the way that government works for the average American.
0
0
4
932
Very cool way to recruit!
The standard graduate application process is backward.
You spend months researching programs, navigating faculty pages, and emailing professors who may or may not respond. You pay hundreds of dollars in application fees. Then you wait to find out if a lab actually has space, t in your research.
Great AI research starts with the right supervisor match. The process should too.
🧵
0
0
2
1,298
Daniel Severo retweeted
Bend 2 is here!
It is a new programming language that blocks AI mistakes via *proof checking* - the same technique big AI labs used to solve open math problems, like Navier-Stokes.
It is also very fast, and runs on GPUs.
Watch the video. Link in the comments.
It is a new programming language that blocks AI mistakes via *proof checking* - the same technique big AI labs used to solve open math problems, like Navier-Stokes.
It is also very fast, and runs on GPUs.
Watch the video. Link in the comments.
RELEASE DAY
After almost 10 years of hard work, tireless research, and a dive deep into the kernels of computer science, I finally realized a dream: running a high-level language on GPUs. And I'm giving it to the world!
Bend compiles modern programming features, including:
- re support
- Unrestricted recursion and loops
- Fast object allocations of all kinds
- Folds, ADTs, continuations and much more
To HVM2, a new runtime capable of spreading that workload across 1000's of cores, in a thread-safe, low-overhead fashion. As a result, we finally have a true high-level language that runs natively on GPUs!
Here's a quick demo:
666
1,331
11,865
2,215,058
The best proxy for happiness I've seen so far is the amount of time you can spend with your kids, conditioned on you enjoying your craft.
0
0
3
643
Daniel Severo retweeted
Last month I wrote about how we can build a positive and safe future for everyone: meta.com/thefutureisfor…
Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens.
The reality is:
- People won't want to use agents that are misaligned with them and that don't do what they ask, so labs have a strong natural incentive to make their models more aligned.
There is a lot of debate about slowing progress on capabilities until alignment catches up. My view is that trust and alignment are quickly becoming the most important capabilities that will differentiate agents and models. Any lab that doesn't focus on alignment will fall behind.
- Labs face significant liability if their models cause harm, so they have a strong incentive to prevent this as well.
Meta delayed shipping Muse for several months to focus on safety and security. We didn't call for everyone else to do this before we would. We just did it as part of our day-to-day work because it was clearly the right thing for people and for us. I'm proud of the security foundations we've built.
- Engaging independent evaluators and advisors is industry best practice. MSL already does this today in several areas because it helps produce better work. Other labs can just do this too. In general, it would be helpful for there to be a larger and more diverse ecosystem of evaluators.
- Committing the significant majority of compute towards serving people rather than racing towards recursive self-improvement is one of the best ways to ensure we develop this technology safely. Meta has made this commitment and other labs can do this as well.
I believe the key to building a positive future for everyone is maintaining the right balance of power. This is within our power to do.
Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens.
The reality is:
- People won't want to use agents that are misaligned with them and that don't do what they ask, so labs have a strong natural incentive to make their models more aligned.
There is a lot of debate about slowing progress on capabilities until alignment catches up. My view is that trust and alignment are quickly becoming the most important capabilities that will differentiate agents and models. Any lab that doesn't focus on alignment will fall behind.
- Labs face significant liability if their models cause harm, so they have a strong incentive to prevent this as well.
Meta delayed shipping Muse for several months to focus on safety and security. We didn't call for everyone else to do this before we would. We just did it as part of our day-to-day work because it was clearly the right thing for people and for us. I'm proud of the security foundations we've built.
- Engaging independent evaluators and advisors is industry best practice. MSL already does this today in several areas because it helps produce better work. Other labs can just do this too. In general, it would be helpful for there to be a larger and more diverse ecosystem of evaluators.
- Committing the significant majority of compute towards serving people rather than racing towards recursive self-improvement is one of the best ways to ensure we develop this technology safely. Meta has made this commitment and other labs can do this as well.
I believe the key to building a positive future for everyone is maintaining the right balance of power. This is within our power to do.
1,920
2,823
28,277
7,796,386
Someone make an RL env benchmark please
super smash bros melee is fully decompiled, this project started in july 2020, so it took over 6 years. the progress was hugely accelerated by llms and finished off with astra. it's a very large game with 3.88mb of compiled code.
github.com/doldecomp/melee
0
0
4
790