Macrocosmos launches IOTA for liquid training on scattered, disaggregated compute
markjeffrey · x · 2026-09-13
- Macrocosmos, a Bittensor ecosystem team, launched IOTA, a "liquid training" platform built on disaggregated training.
- Core idea: instead of one reserved block of matched hardware, IOTA coordinates training runs across scattered, mismatched, constantly changing compute as one unified system.
- It targets three pain points: training economics (train on compute you don't own), capacity constraints (turn idle/mismatched compute into usable capacity), and platform utilization (squeeze more from existing fleets).
- The poster claims Macrocosmos is the first team on Bittensor to make this work.
More from Infra
- DeepMind Chief Strategist: AI Infrastructure Spending Is a Bet on Recursive Self-Improvement — rohanpaul_ai · 2026-09-13
- AI valuations can't all be right: memory at 3-5x PE vs premium infrastructure, says Gavin Baker — rohanpaul_ai · 2026-09-13
- Draw Things launches Local Code beta: local coding agents on Mac at ~980 tok/s prefill — liuliu · 2026-09-13
- Dev launches Lorivo: one GPU server serves many LoRA adapters via vLLM — TheOneWhoWil · 2026-09-13
- "The CUDA moat is gone": Japanese neocloud ai& deploys Tenstorrent at scale — DavidBennett__ · 2026-09-13
- Patched vLLM+FlashInfer Pushes Gemma 4 31B to 150 tok/s on a Single B300, Beating SGLang — abhijithneil · 2026-09-13