Illia (root.near) (๐บ๐ฆ, โ)
@ilblackdragon
Co-Founder @NEARProtocol. LLM co-inventor. @ekang426 husband.
๊ฐ์
October 2009
462 ํ๋ก์ ์ค 118.1K ํฌ
This keeps happening that models seem โdumberโ after initial launch.
Verifiable inference with @near_ai proves that you are using the same model and same code for inference in every response.
๐จ GPT-6 Astra (Launch) Vs GPT-6 Astra (Today)
Same exact prompt
But todayโs result looks noticeably worse
It genuinely feels like OpenAI nerfed Astra
