VQV Signal
According to a post on X by @QwenAIFans, a developer compiled an NPU into the weights of Qwen3.5-4B and used it to run Karpathy's stories260K, a small model that writes the stories shown in the demo video. The post says this follows an earlier experiment by the same developer that put a RISC-V core inside model weights. The account describes the work as an experimental demo, not a usable speed.
According to a post on X by @QwenAIFans, a developer compiled an NPU into the weights of Qwen3.5-4B and used it to run Karpathy's stories260K, a small model that writes the stories shown in the demo video. The post says this follows an earlier experiment by the same developer that put a RISC-V core inside model weights. The account describes the work as an experimental demo, not a usable speed.
AI-assisted summary based on the listed source.
After an earlier experiment that put a RISC-V core inside model weights, the same developer compiled an NPU into Qwen3.5-4B and used it to run Karpathy's stories260K, which writes stories in the demo video. The Qwen fan account notes it's an experimental demo, not usable speed. Original post on X Section:...
For now this looks like an early proof of concept rather than a practical tool, since the post says the speed is not usable. Developers working on open-weight models and on hardware-level inference may find the approach worth tracking. Next, watch whether the developer shares more technical detail or performance figures.
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.
