‹ BackHN Continuity

Thread

The Download: why AI's latest breakthroughs and fears may be more hype than rea

49 points · 128 comments · joozio

  1. Tbarlow · · focus · HN ↗
    Feels like the blockchain bubble all over again, just with more impressive demo reels. Still waiting for my LLM to write perfect code.
    1. ahtcx · · focus · HN ↗
      Do you write perfect code? I was in denial for a long time too, but the reality is this is how software engineering is now. It can handle pretty much any codebase and vastly faster than you ever will be able to. With enough context, it writes good code and it's only getting better.
      1. tomrod · · focus · HN ↗
        > Do you write perfect code?

        For simple things, occasionally!

        For more complex things, usually not, but I write working code. LLMs will sometimes write working code, sometimes not.

        1. ahtcx · · focus · HN ↗
          If you haven't tried any of the newer models from the last ~6 months, you're out of touch. A year ago, LLMs were essentially just a clever search engine for me, I found they fell apart pretty quickly on bigger codebases and more complex projects.

          It doesn't always write perfect code on a first pass but nor do you, and it's amazing at understanding any issues it may encounter, even very low level ones. I don't think there's anything I can do that it can't now. There is a learning curve to working well with LLMs, but once you get there you'll be producing high quality code faster than you can imagine.

          Part of me wishes they weren't so good. I enjoy(ed?) the process of working through problems and it does mean you might no longer have a perfect mental model of your code. But I do also enjoy being able to build things well and fast.

          1. tomrod · · focus · HN ↗
            Opus 5.1 was just so darn verbose and overtuned. I look forward to trying out the new Luna/Sol and recent Claude fixes.
            1. ahtcx · · focus · HN ↗
              Yeah to be honest I still find Claude too verbose with Fable 5.1, I much prefer the output style of OpenAI's models. With strict guidelines it can be improved but it tends to drift in longer conversations.
Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.