4 pointsby masaTokyo5 hours ago2 comments
  • lunaprompts_hn3 hours ago
    One thing that doesn't get discussed enough: the gap between people who can use AI tools effectively and those who can't is widening fast.

    The engineers who thrive are the ones who understand how the underlying models actually work. They know when to trust the output, when to add constraints, and how to structure prompts for reliable results across different contexts.

    The ones struggling tend to treat LLMs as magic boxes. They get inconsistent results and don't know why.

    This is creating a real challenge for hiring managers. Standard coding interviews do not surface which category a candidate falls into at all.

    • CSSer2 hours ago
      The best way I can describe it is not that different from explaining how to effectively use a search engine in the past.
  • TomKwon5 hours ago
    Well great idea. CDP over WebSocket to drive the Electron shell is a clean approach, no reverse engineering needed and you stay within supported boundaries.