Gensyn builds IR3DE-AXL, a decentralized collective inference network with no central gateway
benfielding · x · 2026-09-10
Gensyn's research team unveiled IR3DE-AXL, a prototype that replaces the standard "one central gateway + fixed model catalogue" inference setup with a decentralized collective inference network.
- Core idea: every participant runs an AXL node, acting as a user, an inference provider, or both; each node builds its own router from whatever experts the network currently offers and can contribute its own expert models in return.
- Two components: IR3DE handles local routing — it embeds the prompt, scores tokens against domain statistics via ridge regression, and lets confident token-level votes pick the most relevant expert. AXL is the peer-to-peer layer that turns routing decisions into live cross-node inference requests.
- The post includes full setup, inference flow, and UI walkthrough, and openly discusses limitations of the prototype.
More from Infra
- ByteDance reportedly building 5-6 GW of AI data centers in Ulanqab, costing $110-130B — zephyr_z9 · 2026-09-10
- DeepSeek Raising Second Round at $71B After $52B Round, $500M Revenue Run Rate — zephyr_z9 · 2026-09-10
- Rubin Cuts Inference Cost Up to 10x, 2028 Kyber NVL1152 Will Link 1,152 GPUs in One NVLink Domain — imadade · 2026-09-10
- NVIDIA reportedly paying hefty premiums to lock up probe card and test socket capacity — zephyr_z9 · 2026-09-10
- Day 249 of GPU Programming: Tracking Cerebras From CS-1 to WSE-3 Turbo-Powered CS-4 — blaizedsouza · 2026-09-10
- GPT-6 Astra pre-training cost estimated at $432M, full model $1-2B — scaling01 · 2026-09-10