Last night I learned about NotebookLM and suddenly thought, "I'll try using this to write articles." In the end, I had generative AI draft four articles for me.
- How Angular Signals Made State Management So Much Easier
- Angular v19 Release: Standalone Components by Default and Signals Reach Stable
- Angular 19.2: Experimental HTTP Resource and Response Streaming
- Angular v19 Style HMR: A Dramatic Boost to the Developer Experience
The workflow started by loading an official Angular YouTube video (in English) into NotebookLM and generating a summary. From that summary, Devin drafted the article in my voice; I fine-tuned it in Cursor, then edited it directly myself. The rewrite was maybe around 10%.
The promise of generative AI—and the question it asks
I was honestly surprised by what generative AI can do. What impressed me most was its ability to read between the lines of a 30-second English short and expand it to a target word count. English videos take energy to follow, and video itself takes time to watch, so I had not been watching them actively. Being able to summarize information, shift perspective, and draft an article in about ten minutes per piece feels revolutionary.
Fundamentally, I do not think the value of output like this is "hey everyone, I learned this!" In essence, generative AI trims source data with prompts into an article, so ideally everyone should access the original video and pick up what they need. That might be different if you are publishing from a niche source few people visit—but these are official Angular videos.
What did have value for me was articles generative AI wrote as "me." It takes source material and hypothesizes how I would verbalize it from my past knowledge and experience—"you would do it like this, right?" Some hits were spot on; some missed the mark. It does what would normally require hiring a ghostwriter. I am less interested in writing efficiency than in the model of information processing.
I genuinely think generative AI is impressive here. At the same time, it asks: "Can you really use this well?"
When Hatsune Miku swept the world with "anyone can be a singer and make singing videos," everyone eventually noticed: "but you still need to know how to make music." Generative AI will write for you, but when you read the output it says plainly: "but you still need editorial skill."
Summary
For this article, I wrote about fifteen bullet points first; generative AI filled in context around them; then I rewrote about 80% of it. It is weak at developing an argument with subjective perspective on a phenomenon rather than stating facts—maybe because it does not know where it is headed and wanders.
In the end, accepting output as-is, lightly editing it, or rewriting almost everything—as I did this time—is the operator's right and responsibility.
See you next time.