Content pfp
Content
@
0 reply
0 recast
2 reactions

Gabriel Ayuso pfp
Gabriel Ayuso
@gabrielayuso.eth
The ceiling is set by the dev using LLMs. They won't turn a 1x swe into a 10x swe. A 10x swe without LLMs is still magnitude better than a 1x swe with the best coding LLM with deep devenv integration. However, an LLM could help turn a 1x swe into a 10x swe if they use it to learn. Such as by asking how to improve things and why, by looking at top tier OSS projects and asking the LLM why the dev made certain decisions. I'm doing this with react now since I'm still getting used to it.
5 replies
3 recasts
13 reactions

dusan.framedl.eth pfp
dusan.framedl.eth
@ds8
planning to rewrite nextjs api in rust next week. never touched rust before, let's hope claude knows what he's doing.
3 replies
0 recast
2 reactions

Gabriel Ayuso pfp
Gabriel Ayuso
@gabrielayuso.eth
I had a bad experience with LLMs and Rust last fall. It kept giving me wrong suggestions and I'm not familiar with Rust so it was annoying.
1 reply
0 recast
1 reaction

dusan.framedl.eth pfp
dusan.framedl.eth
@ds8
i expect 1y will make a difference (currently switched from copilot to claude and the difference is mind blowing). will see... worst case i'll just use java
2 replies
0 recast
0 reaction

Gabriel Ayuso pfp
Gabriel Ayuso
@gabrielayuso.eth
Right, I was using gh copilot back then. Should be much better now with claude
0 reply
0 recast
1 reaction

atlasmoth pfp
atlasmoth
@osuji.eth
Curious to see how this goes, please share your progress!
0 reply
0 recast
1 reaction