600+ Agent learning materials open-sourced in one shot. We're devoting space to it because it reveals a more honest problem in the current Chinese AI Agent (intelligent agent) learning market: what's really missing isn't tutorials, but a coherent path.
What this is
GitHub user tangshiyegit (汤师爷) recently open-sourced an Agent learning guide that consolidates 628 programming and architecture articles, 23 WorkBuddy pieces, and 37 Coze pieces into a three-stage roadmap: the first stage has AI take over everyday office tasks on your computer (WorkBuddy is a desktop assistant that can read local files and call tools), the second uses Coze (ByteDance's intelligent agent platform) to build agents that can call knowledge bases and plugins, and the third adds engineering capabilities like databases, caching, and message queues to wire agents into business systems.
The project is hosted on GitHub (MIT license) and an online reader site, with VuePress 2, Vue 3, Vite, and TypeScript on the frontend. The author says he will keep supplementing tutorials on Codex, Doubao Work, and Tongyi Office.
Industry view
Supporters think this is the right move. Agent tutorials exploded over the past year—prompt templates, installation guides, and platform reviews scattered everywhere, with the universal pain point being "individual pieces are useful, but they don't connect when stitched together." Stringing the three stages into one line matters for anyone trying to move from "can use" to "can build."
But we should also flag a few risks. First, Agent tools iterate extremely fast—WorkBuddy and Coze's interfaces and capabilities could shift significantly within six months, which the author himself acknowledges in the project description. Open-sourcing slows obsolescence but doesn't eliminate it. Second, the content is concentrated on specific platforms, creating path dependency; once those platforms shift strategy, the entire roadmap's value takes a hit. Third, 628 Markdown files are a labor of volume—the gap between "follow along" and "truly understand" still remains, and the barrier for true beginners isn't low.
Impact on regular people
For enterprise IT: This kind of open-source path is well-suited as a blueprint for internal Agent training, but watch out for platform lock-in behind any single tutorial—reference it, don't copy it wholesale.
For individual professionals: Non-technical roles (operations, content, HR) can start by picking one repetitive task from this week and trying the WorkBuddy section. But "finishing the tutorial" doesn't equal "knowing how to use AI"—the trial-and-error cost is still on you.
For the consumer market: The Chinese AI training market may shift from expensive paid courses toward high-quality open-source resources, and platforms like Coze and WorkBuddy will likely become more willing to play along with tutorial ecosystems.