Hybrid Linear-Attention Architectures
Open models will keep replacing much of full attention with fixed-size recurrent layers such as Gated DeltaNet, and that fixed state will increasingly be used as an engineered resource for memory, tool execution and cheaper long-context inference.
forming · confidence 40 · Emerging (watchlist) · tracking since October 29, 2026 · updated October 29, 2026
Why the conviction moved
- Oct 29Strengthened +4
A developer's Qwenputer compiles an exact RISC-V CPU into Qwen3.5-4B and uses its Gated DeltaNet recurrent state as working memory, running C and MicroPython inside the forward pass and scoring 97.5% to 100% on DeepMind arithmetic tasks. This shows the fixed recurrent state being used as an engineered resource for tool execution, at the cost of doubling the checkpoint to 18.8 GB.
Source trail
Supporting · October 29, 2026
Developer compiles an exact RISC-V CPU into Qwen3.5-4B and uses its Gated DeltaNet state as memory
A developer's Qwenputer compiles an exact RISC-V CPU into Qwen3.5-4B and uses its Gated DeltaNet recurrent state as working memory, running C and MicroPython inside the forward pass and scoring 97.5% to 100% on DeepMind arithmetic tasks. This shows the fixed recurrent state being used as an engineered resource for tool execution, at the cost of doubling the checkpoint to 18.8 GB.
Hacker News / divagr.com
Unlock full source trail, score history, and daily updates.
Unlock TrendsTownsquare
Argue the thesis in Townsquare.