Pluralis Research
Pluralis Research is a leading voice at the intersection of frontier AI models and Web3 innovation. They provide insightful content on how individuals can train, own, and leverage decentralized AI, exploring the future of technology on platforms like Twitter.
Recent Promotions
No promotions recorded yet.
Promoted projects will appear here.
Latest X Posts
@SameeraRamasin1 (and even more impressive -- @SameeraRamasin1 cooked this paper up in ~2 weeks and just presented it at ICLR. Maybe a record?) https://t.co/WiGBHpMCcn
One of the hardest unsolved problems in AI training: models get surprisingly brittle at scale. A single instability can trigger divergence + slow weight corruption that only shows up checkpoints later — wasting days of compute and millions of dollars. @SameeraRamasin1 architecture warm-up work tackles this. Especially critical for distributed training where you can’t fully control who joins the network or when.
Factored Gossip DiLoCo (by @ChaminHewa) has been accepted to ICML 2026. It removes the all-reduce required to compute the outer-optimiser step, improving robustness to failed nodes. In a collective training setting, this allows nodes to leave arbritarily with minimal impact. https://t.co/UIpWHUat3G
Frequently Asked Questions
Related Influencers
vitalik.eth
vitalik.eth offers profound insights into blockchain technology, the evolution of Ethereum, and the broader Web3 ecosystem. His content delves into decentralized systems, the philosophical underpinnings of digital currencies, and the future trajectory of web decentralization, shaping conversations across various platforms.
Watcher.Guru
Watcher.Guru provides real-time, unbiased coverage of the global cryptocurrency and finance markets. Through high-frequency updates on Twitter, they deliver breaking news on blockchain regulation, market movements, and institutional adoption to a global audience.
CoinDesk
CoinDesk is a prominent global media company delivering news, insights, and data on cryptocurrency and blockchain technology. They cover market trends, future innovations, and host industry-leading events like Consensus, providing comprehensive analysis across various digital platforms, including their podcasts and specialized market reports.
Mario Nawfal
Mario Nawfal is a leading Web3 entrepreneur and host on X, providing 24/7 live streams and market analysis on business, technology, and global crypto markets. As a venture capitalist, he delivers deep insights into startup investing, digital asset trends, and breaking blockchain news.
Gary Vaynerchuk
Gary Vaynerchuk is a leading entrepreneur and NFT pioneer, renowned for founding VeeFriends and shaping the digital asset landscape. His content offers educational insights, market analysis, and business strategies, bridging traditional ventures with Web3 culture to highlight long-term value, community building, and digital ownership across various global platforms.
Cointelegraph
Cointelegraph is a leading global media outlet providing comprehensive coverage of the cryptocurrency and blockchain industry. Since 2013, they've delivered breaking news, in-depth research, and expert podcasts across multiple digital platforms, offering critical analysis and interviews for the Web3 community.
Latest X Posts
@SameeraRamasin1 (and even more impressive -- @SameeraRamasin1 cooked this paper up in ~2 weeks and just presented it at ICLR. Maybe a record?) https://t.co/WiGBHpMCcn
One of the hardest unsolved problems in AI training: models get surprisingly brittle at scale. A single instability can trigger divergence + slow weight corruption that only shows up checkpoints later — wasting days of compute and millions of dollars. @SameeraRamasin1 architecture warm-up work tackles this. Especially critical for distributed training where you can’t fully control who joins the network or when.
Factored Gossip DiLoCo (by @ChaminHewa) has been accepted to ICML 2026. It removes the all-reduce required to compute the outer-optimiser step, improving robustness to failed nodes. In a collective training setting, this allows nodes to leave arbritarily with minimal impact. https://t.co/UIpWHUat3G