Sam Lehman
@SPLehman
Followers
2K
Following
2K
Media
95
Statuses
1K
Investor @SymbolicVC & @0xbeaconcom | Co-Founder of @0xmitharvard | @MIT Alum
Cambridge, MA
Joined December 2021
RT @WhatTheLJW: Grateful to share that I’m joining @psdnai as a founding member and VP of Strategy & Ops. We’re announcing a $15M seed rou….
0
6
0
RT @0xKanekiKen: Attention is the fuel of the internet. Ad slots turn it into dollars but the flow sits locked inside a few platforms, out….
0
5
0
RT @0xKanekiKen: Grateful for the support from @sandeepnailwal @SPLehman and the @NailwalFellows that pushed @0xTokenClick from idea to pil….
0
1
0
RT @TheOneandOmsy: With Robinhood, all you really need to believe is they'll be a primary beneficiary of the $10t+ wealth transfer from our….
0
21
0
RL But Make it Art.
Initializing planetary scale inference.
0
0
16
Democratizing the creation of new gyms/environments, and not just rollouts, was something that really excited me about distributed RL. Really cool to see @gensynai release this!.
1/.Introducing RL Swarm’s new backend: GenRL. A modular reinforcement learning library built for distributed, fault-tolerant training - now powering RL Swarm from the ground up. 🧵
0
1
8
Sounds a little. modular
whoa so @thinkymachines is doing model merging + customized RL. quite a come-up for merging in the past couple weeks, with @arcee_ai mergekit also featuring heavily in AFM. credit due to @jeremyphoward for being the first to make me take modelmerging seriously
1
0
4
Excited to be talking all things deAI on this panel tomorrow!.
Sign up for the AI and robotics day on June 24th here: We'll kick off with a panel on the future of DeAI featuring:.- @SPLehman.- @karsenthil.- @0xalpo. Followed by project showcases from:.- @openmind_agi.- @BitRobotNetwork.- @theveldtai.- @indexnetwork_
0
0
6
Expect a lot of new updates coming from MITxH in the coming months. Big things planned🚀.
1
0
11
RT @FerdinandMom: The research paper video review on "Swarm Parallelism" along with the author @m_ryabinin, Distinguished Research Scientis….
0
9
0
You heard about Golden Gate Claude, now meet Batman OLMo🦇 Very cool open-source work on concept steering from one of the brightest researchers I know, @hbfreed! Read the full report and play around with the demo below👇.
I've spent the last couple months exploring sparse autoencoders (SAEs) inspired by @anthropic's feature steering work. Using @allen_ai’s OLMo-2-7B, I found some really fun concepts hidden within the model including Batman, Japan, and Baseball. Full details and demo below. 👇
1
0
2
RT @0xSehaj: AI agents are coming to @avantisfi. @bankrbot is the first of many external integrations in the pipeline, several teams are….
0
12
0
RT @MattBeton: exo v2 pending…. completely rebuilt from the ground up using event sourcing design principles; scalable, reliable and future….
0
5
0
Sympathetic to the cost tradeoffs w/ decentralization but, here, that sort of misses the point of this specific work. much more about parallelizable training and modularity in model design.
arxiv.org
We present Branch-Train-Merge (BTM), a communication-efficient algorithm for embarrassingly parallel training of large language models (LLMs). We show it is possible to independently train...
0
0
4
I had the privilege of first meeting @0xSehaj when he was an investor and I jumped at the opportunity to back him when he left to start @avantisfi. In a short time, he's built Avantis into the largest derivatives DEX on @base with over $7.5b in volume. With this raise and their.
We’re excited to announce our $8M Series A, co-led by @PanteraCapital, @foundersfund. This raise accelerates our vision to build DeFi’s universal leverage layer — for everything. Gold, FX, Oil, Equities, Crypto, Sports and more: 24/7, fully onchain, and accessible to anyone.
1
0
12
RT @Ar_Douillard: Really impressive work on compressing activations & gradients for pipeline parallelism. * Constrains projection matrices….
0
8
0