Something that's been bugging me about the AI rollout, with everyone talking about like non-deterministic adversarial multi-agent workflows. Like how do you teach normies about this? Especially, how do you teach normies about this without a traditional path to understanding the underlying mental models?
And like particularly - one thing I talk to a lot of engineers about is what I call "merge days". Merge days are when you like take on the mental overhead of actually getting code merged and deployed, because maybe you have to confirm that tests are run, and you have all the right tests, and like. That mental load on like 50k lines of code is just bananas.
Part of the point is, as has been spoken about forever - actual value is not measured in lines of code. While I think there will be some who can really super charge things with these new powers, I'm not sure how we'll actually scale this.
Like, most of the superpowers in my life has come from single-threading and deep thought. I don't think AI fundamentally changes this, I suspect for most people, it makes delivering real value harder.
