Knowledge/Process collectors

[TomLucidor]

From the following comment https://www.reddit.com/r/vibecoding/comments/1o547xp/comment/nj6xyem/

I simply ask perplexity "labs" to create me a learning dashboard filled with use cases, how to guides, what to do in different scenarios, and where AI tends to fuck up in regards to [topic I'm researching]. Then, I use a notebookLM filled with high quality content (books, etc.) on the coding language and have it create a series of reference guides my LLMs can reference when needed. Each is designed to provide just enough instruction without overloading context.
This is how I taught myself github, docker, then containerization as a concept, then middleware, then caching, then etc... You can teach yourself just about anything 10X faster with AI... Part of me surprised people aren't just teaching themself to code instead, as I just started doing that and i'm amazed at how much I'm picking up. Next I want to teach myself Godat for game development.

I wonder if Open Notebooks + knowledge caching for best practices are enough to get people started.

[TomLucidor]

From another thread, and the "deep work" thing seems effective even if it sounded gauche https://www.reddit.com/r/vibecoding/comments/1o76qvc/comment/njm0hqf/

I totally like the way Codex executes things in a kind of "deep work" mode, where it goes through each step, creates multiple files, shows some implementation steps, it's nice and convincing.

P.S. And also "Rubber ducking" https://www.reddit.com/r/programming/comments/1kyyxgg/comment/mv1w52j/