synchopate

synchopate is an individual GitHub developer account active since December 2020. It publishes open-source software focused on running large language models on CPUs. Its best-known project is llama-modchip (Apache 2.0), an inference engine that runs Mixture-of-Experts models — including the 284B-parameter DeepSeek-V4-Flash and 30B-class models — on machines with 16–64 GB of RAM without a GPU. The account also publishes ARC-AGI and CyberGym benchmark materials built around a Crystalline cognitive memory layer. No real name, location, or corporate affiliation is publicly disclosed.