| ▲ | lavasalesman 2 hours ago | |
Challenge is not gone. Use LLMs to clear up your backlog of things you wanted to do before but couldn't, then choose something that's challenging even with the help of an LLM. Dive into a skill you thought you didn't have time to learn. Make a something you previously thought was too big to make by yourself. | ||
| ▲ | criddell an hour ago | parent | next [-] | |
I’ve been trying to get better at analyzing performance traces (like Bruce Dawson has written about on his blog), crash dumps, and assembly language in general. AIs are still pretty bad at all this stuff. It’s made we wonder if future AIs will get much better at machine language and will skip writing human readable source code and just dump out CPU opcodes (ie write the binary directly). | ||
| ▲ | 0c3ca83 an hour ago | parent | prev | next [-] | |
What I want to do is think about the challenges in my backlog. | ||
| ▲ | amelius an hour ago | parent | prev | next [-] | |
And help the AI labs to become good at it ... | ||
| ▲ | gib444 an hour ago | parent | prev [-] | |
This. Running a load of things on a server at home was on my backlog for ages. I started manually then got Claude to reverse engineer the system and create a load of Ansible that we iterated on. It still needs its hand holding , and likes to pretend it doesn't know what good Ansible looks like but we got there It also wrote a nifty remote LUKS unlock utility. So handy | ||