Developer Builds a Working CPU Inside an AI Model's Memory
An independent developer built 'Qwenputer,' a deterministic RISC-V CPU that runs entirely inside Qwen3.5 by using the model's Gated DeltaNet layers as RAM. It can execute arithmetic and run real C and MicroPython code without any external runtime or tools, and even includes experimental in-model file storage. It is a research curiosity showing how far models' internal memory mechanisms can be repurposed, not a practical product.
Why it matters
๐ป Developer ยท A wild proof of concept showing you can repurpose a model's internal memory layers as literal computer memory; worth a read if you're curious how DeltaNet layers work.
๐ฆ Product ยท Not product-ready, but a sign of how creatively people are exploiting model internals, something to watch for future tooling ideas.
๐จ Design ยท Not relevant to design work.
๐ Business ยท No near-term business impact, but it's the kind of hack that sometimes previews real technical breakthroughs a year or two out.
๐ค Just Curious ยท Someone built a working computer that runs entirely inside an AI model's 'brain,' using its memory as RAM, and it can actually run code.
Sources: Qwenputer: a fully functioning, deterministic RISC-V CPU inside Qwen3.5