[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: computergunfrog.jpg (48 KB, 975x975)
48 KB
48 KB JPG
>use openai api to have chatgpt generate 64 versions of the code you want and pick the winner in a march madness style tournament

why are vibe coders not doing this? are they just too lazy?
>>
>>107910187
Who's gonna pick the winner? You wanna do 64 code reviews on the same feature?
>>
>>107910255
are you retarded? i said use the openai api

>round 1
>32 pairs of code files
>for each pair, have the prompt be something like "you are a code reviewer, your job is to pick the better one"
>>
>>107910265
>are you retarded
The proceeding information you gave after that very clearly shows which one of us two is the retard. Always a frogposter.
>>
>>107910187
Vibe coders should be picking the best out of 64 methods to rope themselves.
>>
>>107910265
>trusting AI to actually test the code
>>
>>107910187
It's fun to think of creative uses of LLMs.
Generally when you use AI to code, you don't just make a single API request and expect it to output the code all at once. You use an "agent" which results in a long series of messages with tool use requests, tool use responses, reasoning blocks, etc. It greps through the codebase, makes changes to various files, reviews compiler errors and linter warnings...

You could do that 64 times in parallel, with 64 different instances of the entire development environment needed for the project (probably in containers) and then judge them. It would be complex to set up, and obviously expensive.
>>
>>107910187
- Generate 2048 different decision trees
- Fuse the predictions of all trees
- Results outperform SOTA neural net model
> too lazy
No, they're too poor. 64x the troughput is 64x the cost. Also, I assume this is what the services do on the backend already.
>>
>>107910187
This is effectively what Gas Town is

It is extremely expensive right now. If it takes 50k tokens to complete a coding task with one agent, you now need ~5M tokens
>>
>>107910187
they already do this, just one at a time
>>
>>107913361
Yeah in my experience it just bullshits you
>u sure this last change won't fuck anything up?
>SURE I TESTED 2000 SCENARIOS IT'S PERFECT
>try it out myself
>immediate crash
>>
>>107916596
>oopsy whoopsy I did a fuckie wuckie tehe
>>
>>107916596
well yeah of course it never tested it it's a text predicting chatbot
ask your toaster next time to run your code
>>
>>107910187
i do this, but not until 64. less then 10 and not using chatgpt.



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.