@imjameshall This is one of those advancements that look trivial at first blush but, in reality, is going to have repercussions that we haven’t even imagined yet.
@jamonholmgren It's a new mode of working certainly.
I still think you can feel the code pushing back when doing AI-assist first, but it manifests differently.
- Changes to features always have a high defect or hallucination rate
- The produced work doesn't quite pass the "eye test"
@thekitze Hard same. I always got the best highs from seeing the positive outcomes that software I built would give to people.
The puzzle solving bit can be fun, but it was never "the main thing" for me.
@lucas_montano I get the vibe that what you're describing is less about coding itself and more about entering and staying in a flow state in the face of a hard problem.
I think that is still very much possible in a software development job assisted by LLM's, and in different, enjoyable ways.
@niall_obrien@wesbos For these kinds of exercises, I’d personally lean into how folks use AI assist for a small product spec and talk through how they think about systems.
It’d have to be live though, as take homes are very easily gamed these days.
@wesbos Overall, yes. I happened to have a side project I was already working on, so show and tell was straightforward. Not all folks have something to show though so YMMV.
I think traditional DSA and technical trivial pursuit rounds should be extinct.