Hi, I’m David Zhang 👋

AI moves too fast for tech media, which all sounds the same. Twitter is too fragmented to remember anything. Long-form articles take forever to get through. Official blogs are just product launches. The only way a normal person can keep up with AI is to listen to the people building it talk. Here are the five Chinese and five English podcasts I think are worth the time, plus five ways to actually finish the English ones if you’re not a native speaker.

Most Obsidian-as-second-brain tutorials skip the part where it breaks: once the vault gets big, Claude Code’s index-based search starts to fall apart. QMD, a local RAG tool by Shopify CEO Tobi Lütke, fixes exactly that gap — hybrid retrieval (BM25 + vector), fully local, free, and a 96% drop in token use. This post covers when to install it and the one prompt that gets it running.

Rie Kudan won the Akutagawa Prize, Joe Vasicek cut his book cycle from 2 years to 14 days, Leanne Leeds ships 12-15 books a year at a 4.5-star Amazon average, and Hachette pulled a novel after Pangram clocked it at 78.4% AI-generated. The Sudowrite / NovelCrafter / Claude + Scrivener stack is now the Western mainstream, and the Chinese-language world has nothing equivalent.

Raw models are no longer enough for serious engineering work. What actually separates mediocre results from great ones is the Harness built around the model: context management, tool execution, task orchestration, feedback loops, and architectural guardrails.

A Skill isn’t a longer prompt — it’s a triggerable, executable, testable, maintainable workflow. From what a Skill actually is, to on-demand loading, tool boundaries, model selection, progressive disclosure, validation loops, and how to share it with other people — the full playbook for writing industrial-grade Skills.

After a year with Claude Code, I distilled my workflow into 7 reusable principles. The biggest beginner mistake isn’t lack of skill — it’s treating Claude Code like ChatGPT. Used right, it’s your super-engineer. Used wrong, it’s a hallucinating, lying slacker.

Explores the limitations of using LLMs as simple chatbots and introduces the Model-Context-Protocol (MCP), detailing four essential tools that provide context, planning, and memory to transform AI into a powerful, autonomous engineering partner.

Clarifies the distinction between Tools (primitive functions), Skills (complex workflows), and the Model Context Protocol (MCP) in agentic AI, providing an architectural framework for developers to choose the right abstraction for their use case.

Provides a practical guide to implementing a persistent memory system for AI coding agents, advocating for a simplified, file-based approach with a manual ‘/save’ command to capture session context, delivering most of the value with minimal complexity.

Analyzes the viral success of OpenClaw, an open-source AI assistant that operates locally with OS-level execution, exploring its key innovations like proactive tasks and a decentralized skill ecosystem, while also highlighting its significant security implications.