Content
@
0 reply
0 recast
2 reactions
Gabriel Ayuso
@gabrielayuso.eth
Bundle to reduce context switching. If you're making some changes to a piece of code to make a specific improvement or bug fix. It can be very productive to tackle other improvements and fixes on that part of the codebase
2 replies
0 recast
20 reactions
agusti
@bleu.eth
cursor: fix everything wrong on my code thx
1 reply
0 recast
5 reactions
dusan.framedl.eth
@ds8
*deletes all*
2 replies
0 recast
3 reactions
agusti
@bleu.eth
this happens so often to me, it has me wondering if i should be even coding or not ππ
1 reply
0 recast
1 reaction
dusan.framedl.eth
@ds8
never stop π«‘ but maybe tone it down a little? claude's been lately suggesting me to prepend "you bleu it again" to all my error messages
1 reply
0 recast
1 reaction
agusti
@bleu.eth
πππ I've moved from claude to o1 lately, its hard to keep up with all the AI ngl. thinking on making my own code buddy agents tho, its triing to explain each AI your preferences each time on settings (like use functional react, use latest libs, etc)
1 reply
0 recast
1 reaction
dusan.framedl.eth
@ds8
still in cursor? in what ways is it better?
1 reply
0 recast
1 reaction
agusti
@bleu.eth
not inside cursor no, mostly using chatgpt UI to use its reasoning capabilities, I like the new thinking mode which is similar to the Reason-Act agents im building for social media, and basically iterate over loops to do -research-, planning, and then execute on your ask. I also think probably the higher 128k context windows make this more feasible as the AI/llm can operate with more -data-/context to fulffill my request. AKA: you can ask for a full site + code and be detailed etc. cursor agent is a bit fleeby imho you have to keep nailing the context you pass it down to etc for it to work reliably, finds it gets stuck some times in old versions of my files and rewerites them (after I fixed manually) already so not great
1 reply
0 recast
1 reaction
agusti
@bleu.eth
for example until 01 thinking it was really hard for me to just pass 2 repo's links to the LLM and ask to merge it, and yet with o1 seems doable for the LLM to research both codebases and make reasonings about it on one run
0 reply
0 recast
1 reaction
