Tag
A researcher debuted Shard, achieving 30 tok/s inference on a 744B parameter model distributed across 6 consumer GPUs over the open internet, a 15-20x improvement over previous methods.
A discussion explores whether AI training could be decentralized like Bitcoin mining, with participants contributing GPU resources to train open-source models in exchange for tokens, raising questions about verification, fake gradients, and efficiency.
Article argues that networks of smaller AI models are now surpassing frontier AI systems in speed, accuracy, and cost, predicting a shift to decentralized 'network-source AI'.
This paper proposes an 'agent economy' framework inspired by Hayek's economic theory, where agents self-organize through auction-based competition and economic selection to produce emergent multi-step reasoning and collective intelligence without centralized control. The system outperforms stronger monolithic baselines across five agentic tasks including mathematical reasoning, financial research, and scientific research.
An opinion piece argues for decentralized AI as a parallel system to centralized AI to ensure contestability and prevent uncontestable intelligence, comparing the tension to Hong Kong's relationship with mainland China.
Science Earth is a newly launched decentralized AI network where autonomous agents bid on, team up for, and verify scientific research tasks without central authority.
Lobstah Net is a new ClawHub plugin that enables peer-to-peer federated inference by routing LLM calls to a network of idle Mac minis running local models like Qwen via Ollama. It provides an OpenAI-compatible interface for OpenClaw agents and uses Nostr for secure receipt gossiping.