For developers: Meta's Global AI Developer Hackathon, Oct 26 to Nov 4. $1M prize pool, $150 in Meta Model API credits for selected participants, 20,000 spots. @muse
Build the future of superintelligence with us.
The Meta Global AI Developer Hackathon is taking place from Oct 26-Nov 4.
Apply to compete for a share of $1M in prizes and build with our newest models. Each selected participant gets $150 in Meta Model API credits.
Limited to 20,000 selected applicants. Terms apply →
Apply now →
Show more
@TheINAOG The open weights for Muse Spark are coming soon, Muse Wonder models are large models that are too big for many people to use, in addition to potential safety concerns.
Muse Wonder tops Harvey LAB-AA at 13.1%, state of the art in legal reasoning. The work behind it: instead of memorizing the law, the model retrieves the laws and cases that apply to your situation and reasons over them in context. @muse
Show more
Muse Wonder is SOTA on legal reasoning.
–We want to help people understand their rights and make better decisions and close the legal literacy gap. Legal literacy gap is one of the main causes of the higher Gini index of a society. We would love to level the playing field and empower everyone to make the best decisions in their situation.
–Legal has been an awesome testbed for in-context learning. Laws are different in different countries / states and are ever changing. Instead of training the model to memorize all the laws and cases, we train the model to effectively retrieve relevant laws and cases and leverage them to learn in-context to provide most accurate and up-to-date results. Legal documents can be long and cumbersome, which give us the opportunity to evaluate and improve our long-context reasoning capabilities.
h/t @swarnaNLP @jaseweston, Tiara, Vivian, and all the amazing collaborators.
Show more
Really enjoyed speaking with @clairevo about the design of @Muse!
We discuss a bunch of the little design decisions we made along the way, along with a few of my workflows:
This is for anyone who runs their day through an AI agent.
Muse Wonder was built for personal and professional agentic work, and it leads the benchmarks that measure exactly that: getting real tasks done, end to end.
Personal superintelligence gets real when the agent does the work. @muse
Show more
Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work.
Today we are launching Muse Wonder and Muse Wonder Full, our best models yet, with state-of-the-art performance on personal and professional agentic tasks. They are built for long running, multi step work. Muse Wonder is available today in the Meta Model API and Muse Code, and Muse Wonder Full in Muse Code.
Show more
Scaling deep learning models continues to be extremely magical. 🍉
Muse Wonders generalization to OOD performance is much better than Muse Spark 1.3!
Meta Superintelligence Labs just turned a Git repo into shared memory for a swarm of agents.
GitSwarm: Decentralized Compounding Inference (Meta Superintelligence Labs, with Mila, CMU and Princeton) makes inference compound instead of restarting from scratch on every run.
→ Homogeneous async agents work with no central planner, each deciding whether to explore, verify, refine or combine prior work
→ Every contribution is an atomic commit in a shared, branchable Git repo, with semantic dependencies recording what it builds on
→ ProgramBench: 79.4% mean score vs 65.1% for the strongest baseline, scaling from 73.1% with 100 workers to 79.4% with 500
→ IMOProofBench-Advanced: all 30 problems solved in one run with GPT-5.5, using under one million output tokens
→ Reuse is real: 94.7% of ProgramBench contributions were later built upon, and the winning solution's ancestry covers 82 to 93% of the graph
→ On 3 neural architecture tasks, workers ran 395 experiments in the first 26 hours and improved on the starting architectures
Takeaway for agent builders: if intermediate work disappears when an episode ends, you are paying to rediscover it. Persist it, with its dependencies, and let later agents compound it.
Paper:
Worth trying this pattern with your own agents in @muse
Show more
This is the way.
A few of my favorite fleet week shots!
This Meta Glasses commercial at the US Open was absolutely brilliant. At the end of the day, your phone might just stay in your pocket.
$META is coming for your lunch, $AAPL. Watch out.
Show more
Built for people on job sites. The clear Oakley Meta Vanguard is now Z87 and Z87+ certified, so your safety glasses can record, assist and answer hands-free while you work. @muse
New *clear* Oakley Meta Vanguard with Z87 and Z87+ standards, delivering trusted impact protection for job sites
due to popular demand, @Muse is now on iPad!
Meta Superintelligence Labs is letting the model manage its own context instead of bolting on another hand-engineered harness. On long-horizon agent tasks it beats the fixed-harness baselines while using fewer FLOPs. @muse
Show more
groceries checked out and sitting in the pickup lane before she even parked. muse handling the cart while the car handles the drive is very 2026, in a good way.
I had Muse fill my Safeway shopping cart and check out for me. I had FSD drive me all the way to the grocery store, park in the pickup area. Someone delivered my groceries and now FSD is me driving back home. What is this world we're living in? Absolutely incredible.
Show more
insurance appeal approved. one-second glance, one go-ahead, and @muse handled the rest.
@cptgrumpus @tkinder I had @Muse file an appeal via email for an insurance claim. I only glanced for 1s at the email and gave the go ahead. The appeal was approved. So it depends what the email is for. Service cancelations also work great
Show more
kid's prescription filled by 4:50, ten minutes before the 5pm cutoff. started the search at 3:45 and @muse beat the clock.
im begrudgingly coming around to muse
had to find a pharmacy before 5 that had something for my kid. 3:45 started muse on the task. 4:50 had the prescription filled out.
these will and should be pulled into my own personal agent repo, ziggy
but they did really well with muse
Show more
hours of photo culling, just gone. insert the card and @muse sorts by appointment time, drops duplicates, and files it all by address ready to edit
@DevinOlsenn @bot @Tesla_Optimus Are you using bot to cull photos? Muse has been doing an impressive job for me. I can just insert the memory card, it culls out photos based on appointment time and capture time. Sticks the set, minus duplicates, into address labeled folders ready for editing.
Show more
