Todays Local #LLM / #AI Inference experiment; Running a 753 Billion parameter model hooked up to a local coding agent.No, I've not come into a couple of money trucks full of cash, I'm using an old Dell T5810 with 256GB of RAM (bought well before the memory price spike), a 512GB NVMe SSD, a quantised version of GLM 5.2, and Colibri (https://github.com/JustVugg/colibri)Yes, it's slow, but I'm well past the belief that all agentic coding needs to be interactive, and I'm not hooked on dopamine hit cycles.
Related
Hashtag Jakarta EE number 346 is out!Read about Jakarta EE at JavaZone, Jakarta EE 12 progress, and GitHub Copilot SDK f...
Hashtag Jakarta EE number 346 is out!Read about Jakarta EE at JavaZone, Jakarta EE 12 progress, and GitHub Copilot SDK for Java with Jakarta EE 11#JakartaEE #EclipseFoundation #Com...
Israeli PR wants to answer your ChatGPT questionshttps://www.politico.com/newsletters/politico-influence/2026/08/14/isra...
Israeli PR wants to answer your ChatGPT questionshttps://www.politico.com/newsletters/politico-influence/2026/08/14/israeli-pr-wants-to-answer-your-chatgpt-questions-01038138French...
New blog: Defending Against "Model Collapse" — an overview of how recursive training can cause AI degradation and the st...
New blog: Defending Against "Model Collapse" — an overview of how recursive training can cause AI degradation and the strategies researchers propose to prevent it. Essential readin...