Ask HN: How do you maintain a state of flow when "vibe engineering"?

I use vibe coding for trivial stuff that otherwise will take me lot of time to code myself or for languages I don’t know. I either feed the LLM the whole code module, or since the beginning of a vibe-coded module, I copy all the prompts and responses into a word file (easier to parse) and feed it back to the LLM if I need a bug fix or change. Most of the modules, however, I write myself.

In a given day, you're getting as much work done while also putting in solid YouTube and HN time. I'd say AI is making good on its promise.