| ▲ | black_knight 20 hours ago | |
It is fascinating. Because sometimes my experience is like yours, but other time, like today, I am like “I can’t believe how stupid Opus 5.5 or even Fable 5.2 is”. They will make horrible design decisions, overcomplicating. I will correct and they are like “Of course, your solution is much simpler and better than mine”. And then they are simply boldly wrong at times. About mathematical facts. That’s why we need formalisation when working with them. They are still stupid and can’t be trusted. But theyvare also smart and can come with amazing insight. It’s a tool, but definitely one which cuts both ways. | ||
| ▲ | jfyi 11 hours ago | parent [-] | |
I think it's funny that people gauge their personal experiences with AI against something like this. Do we even know what Alpöge had at his disposal? I think most people underestimate what working with an agent swarm even with the current public technology would be like. We can't assume he didn't have something more advanced either. I don't feel confident my usage of AI on a personal level reflects any of what has been done here. | ||