Anonymous
6/26/2025, 7:46:16 PM No.105712936
Is no one talking about cursor maxxing on here? Where are the ai assisted programming threads?
Replies:
Thread 105712936 - /g/ [Archived: 804 hours ago]
Anonymous
6/26/2025, 7:51:01 PM No.105712978
Good question. Can you use it to do bugfixes and pr on open source github projs and farm reddit gold there?
Anonymous
6/26/2025, 7:51:10 PM No.105712981
>>105712936 (OP)
this board is full of nocoders
this board is full of nocoders
Replies:
Anonymous
6/26/2025, 7:52:12 PM No.105712991
>>105712936 (OP)
only nocoders get any use out of that tranny shit
only nocoders get any use out of that tranny shit
Replies:
Anonymous
6/26/2025, 7:55:53 PM No.105713049
>>105712936 (OP)
nocoder trash, delete your thread
nocoder trash, delete your thread
Replies:
Anonymous
6/26/2025, 7:59:36 PM No.105713075
>>105712991
>>105713049
>BUZZWORD BECAUSE THE COMMENT ABOVE YOU USED IT
SAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAARRRRRRR
>>105713049
>BUZZWORD BECAUSE THE COMMENT ABOVE YOU USED IT
SAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAARRRRRRR
Replies:
Anonymous
6/26/2025, 8:02:26 PM No.105713107
Anonymous
6/26/2025, 8:06:40 PM No.105713141
>>105712936 (OP)
Why do they all use hexagon cube logos?
Why do they all use hexagon cube logos?
Anonymous
6/26/2025, 8:12:31 PM No.105713217
>>105712981
lmao what gave it away?
>>105712936 (OP)
pair this with claude code and it changes how you develop. you're pretty much a project manager that oversees a fleet of shitty junior to mid level coders.
lmao what gave it away?
>>105712936 (OP)
pair this with claude code and it changes how you develop. you're pretty much a project manager that oversees a fleet of shitty junior to mid level coders.
Replies:
Anonymous
6/26/2025, 8:28:35 PM No.105713415
>>105713217
Yeah that's how I approach it now, too. Cursor has essentially turned software development from being mostly about writing code to being mostly about technical project management.
In truth I despise it. I LIKED programming. I don't want to be a bot herder.
Whatever. I guess I can still code for pleasure in the evenings and weekends, at least.
Yeah that's how I approach it now, too. Cursor has essentially turned software development from being mostly about writing code to being mostly about technical project management.
In truth I despise it. I LIKED programming. I don't want to be a bot herder.
Whatever. I guess I can still code for pleasure in the evenings and weekends, at least.
Anonymous
6/26/2025, 8:36:10 PM No.105713481
>>105712936 (OP)
Cursors were already maxxed 30 years ago.
Cursors were already maxxed 30 years ago.
Anonymous
6/26/2025, 9:01:14 PM No.105713718
Extremely useful but there's not much to talk about. Everything is self-explanatory. When people talk about "agentic prompting techniques" it makes me want to rope
Anonymous
6/26/2025, 9:27:31 PM No.105713951
>>105712936 (OP)
>shit already exists in vscode
>you're unemployed or your job is baby tier and solves no complex problems
cool thanks great thread you get paid for it?
>shit already exists in vscode
>you're unemployed or your job is baby tier and solves no complex problems
cool thanks great thread you get paid for it?
Anonymous
6/26/2025, 9:37:19 PM No.105714053
>>105712936 (OP)
I actually like the idea of putting constraints on LLMs coding, Like use MCPs (documentation injected into context in a concise manner), forcing it to use the linter and type constraints so it can't hallucinate methods or functions, force it to work with tools rather than naively appending the whole context of your codebase (using regex and cached definitions for your features; future lookups), using a PR review process to catch best practice and make it more readable (it's more accurate at pointing out errors when asked to criticize the code then to void them when generating), creating test cases systematically so it's a hard one way valve where it can't easily, break previous functionality. It's an interesting engineering problem over creating useful tool-calls to extract the most "effective" reasoning out of the models given context constraints.
I actually like the idea of putting constraints on LLMs coding, Like use MCPs (documentation injected into context in a concise manner), forcing it to use the linter and type constraints so it can't hallucinate methods or functions, force it to work with tools rather than naively appending the whole context of your codebase (using regex and cached definitions for your features; future lookups), using a PR review process to catch best practice and make it more readable (it's more accurate at pointing out errors when asked to criticize the code then to void them when generating), creating test cases systematically so it's a hard one way valve where it can't easily, break previous functionality. It's an interesting engineering problem over creating useful tool-calls to extract the most "effective" reasoning out of the models given context constraints.
Anonymous
6/26/2025, 9:49:58 PM No.105714181
Its so so
Anything not being completely standard stuff like nodejs express, the models have big problems with, spouting out nonsense non existing functions and stuff, you really have to double check everything they put out
Its good for implementing repeating patterns or basic structure like an advanced auto complete, sometimes i use it like a pair programmer, discussing implementation details with it which works well
Anything not being completely standard stuff like nodejs express, the models have big problems with, spouting out nonsense non existing functions and stuff, you really have to double check everything they put out
Its good for implementing repeating patterns or basic structure like an advanced auto complete, sometimes i use it like a pair programmer, discussing implementation details with it which works well