Observed arrival · 2026-09-28
Pooled turns spare devices into one model runner
Pooled presents a browser-based demo for splitting open-model inference across the memory of devices in a shared room.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People exploring shared-device inference and coding agents
- Worth noticing
- The example assigns 40 model layers across three devices and shows about 10 KB moving between them per word.
Field notes
The room flow uses a short code to bring devices together, then displays each device’s memory contribution and assigned layer range. The example lists a 35B MoE model at 22.5 GB and reports a three-device run at 21.4 tokens per second; those figures appear as demo output, not independently verified performance. In Code mode, the agent’s file operations and local preview are visible in the same interface.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue