8 hours. That’s how long the new version of Zhipu AI’s (aka ZAI) flagship model, GLM-5.1, is capable of working autonomously and still cranking out high quality results. ZAI trained ths new model entirely on Huawei Ascend chips, not an NVIDIA GPU in sight. This is China’s leading chip maker for AI, and the benchmarks…
Summary In today's episode, I explain why humans mine topics more effectively than AI to identify high-intent content ideas. Here's what this means for you. You can move beyond generic AI outputs to create highly specific content that addresses actual customer needs. You'll also learn these concepts: how to mine social media and Q&A platforms…
Pyramids to obelisks. Take a look at the job description in the attached image, for an account coordinator at a PR firm. Ask yourself honestly how much of this job AI could do, with today’s capabilities, especially with agentic tools like Claude Cowork. The answer is almost all of it. Creating first drafts of anything?…
Summary In today's episode, I explain how to use first principles to solve the problem of conflicting AI instructions. Here's what this means for you. You can significantly reduce AI hallucinations and improve model performance by establishing a clear hierarchy of rules. You'll also learn these concepts: why conflicting instructions confuse models, how to create…
Agents need job descriptions. If you’re building with true agentic frameworks – OpenClaw, Claude Managed Agents, Hermes Agent, etc. – you’ve likely noticed that getting an agentic system in place can be challenging (to say the least). Configuration alone is kind of a pain – it works best in its own environment, like a Docker…
Summary In today's episode, I explain how to move beyond generic AI responses to unlock deep personal growth and learning. Here's what this means for you. You can transform AI into a highly effective personal assistant by providing it with specialized data and structured memory. You'll also learn these concepts: constructing local knowledge graphs for…
A gentle reminder that the only GUARANTEED private AI is with models you are hosting yourself, on your machine or your infrastructure. Every cloud provider, without exception, will have somewhere in their Terms of Service that they will comply with lawful requests for data from the government of their jurisdiction. For example, this is inference…
Summary In today's episode, I walk through a local AI model roundup covering the latest open-weight models you can actually run on your own hardware. Here's what this means for you. You get a practical playbook for picking the right model, server, and harness combo so you can stop burning through cloud tokens and run…
Summary In today's episode, I explain how to safely grant AI agents command line access without risking your entire file system. Here's what this means for you. You gain the freedom to experiment with autonomous agents without fear of losing your important data. You'll also learn these concepts: how Docker containers provide a safety net,…
