Red Rozenglass on Nostr: Interesting... How much did you spend on tokens for that, I'm curious. I can't seem ...
Interesting... How much did you spend on tokens for that, I'm curious.
I can't seem to get my LLMs to do anything right, like an OpenXR Vulkan VR example that works (even after more than a dozen back and forths), or a correct safe implementation of an encrypting file chunker with content-defined chunking as part of a backup system (it misses the most obvious security issues with the encryption, and never got the chunking correct), or heck, even a snake game in SDL is a bit too much most of the time; they write it, but there's always something about it that makes it not a correct snake game. I tried half a dozen self-hosted ~14b LLMs that are coding-specialized, I tried OpenAI's GPT-4o, and DeepSeek's R1 deep thinking, it was a miserable experience in all cases.
Not sure if I just haven't used "the right" AI, or if asking them to write C is just too much, or what...
I can't seem to get my LLMs to do anything right, like an OpenXR Vulkan VR example that works (even after more than a dozen back and forths), or a correct safe implementation of an encrypting file chunker with content-defined chunking as part of a backup system (it misses the most obvious security issues with the encryption, and never got the chunking correct), or heck, even a snake game in SDL is a bit too much most of the time; they write it, but there's always something about it that makes it not a correct snake game. I tried half a dozen self-hosted ~14b LLMs that are coding-specialized, I tried OpenAI's GPT-4o, and DeepSeek's R1 deep thinking, it was a miserable experience in all cases.
Not sure if I just haven't used "the right" AI, or if asking them to write C is just too much, or what...