KernelKind on nostr

Recent Notes

At this particular stage in the AI landscape, if you're using any other cloud model other than DeepSeek V4.1-Flash . You are definitely losing money. It is so high in quality and so cheap in cost, I c...

KernelKind @KernelKind 1789780063

hmm in your experience does it seem better than its AA rating? glm5.3 flash is cheaper on openrouter & has better score

ABH3PO · 5d

I've used both glm-5.3-flash and deepseekv4.1-flash and have been using them ever since each of them came out, deepseek is definitely cheaper, and get's the job done always more accurately, thinks 2-3 orders ahead on any tasks. maybe you're comparing v4-flash.

𝕞ptf · 5d

i use astra med for some code checking so far. it is quite good admittedly... also great for images the new gptimage 2.5

KernelKind @KernelKind 1789779751

max burns so fast though, ill prob cut back to high. i feel like for some things med flubs a bit

jb55 · 6d

not only did my agent plan this whole sequence, but it prompted each new session itself recursively. i have fully automated my work nostr:note1yhrtl0ccvvyx75m3stqxx9ysltvt2tyxmj2djl7a3djydpcqz4ksfjjl...

KernelKind @KernelKind 1789739488

woah

🤙1

𝕞ptf · 5d

what's your fav AI model right now? I've been testing out poolside/laguna-s-2.1 . It's not the smartest but its free, what can I say. #AI https://blossom.mypathtofire.de/87456372db0f8d9de9f71a4e1a4...

KernelKind @KernelKind 1789739291

astra max

1❤️1

𝕞ptf · 5d

i use astra med for some code checking so far. it is quite good admittedly... also great for images the new gptimage 2.5

Karnage · 1w

Astra absolutely dumb today. Keeps trying to do 1 thing. I ask it if its really the best approach, it says no. Then I say, let's do this other approach. Then it goes down the wrong road AGAIN. THen I ...

KernelKind @KernelKind 1789478978

xD been there

Laan Tungir · 1w

nostr:nprofile1qyxhwumn8ghj7mn0wvhxcmmvqy2hwumn8ghj7mn0wd68ytn00p68ytnyv4mz7qpq6g4umvwj2pduqc8kt2rv6heq2vhvtulyrsr2a20d4suldwnkl4hq62etsy

KernelKind @KernelKind 1789478428

me too, reccomend

KernelKind

@KernelKind 1789181763

thank you @PayPerQ , yall actually have a good product 🙏

i used to use nano-gpt occasionally but its genuinely unusable bc its so buggy now. ppq has officially replaced it in my anonymized llm workflow
r

1❤️2

PayPerQ · 1w

Thank you! It means a lot.

Karnage · 2w

Astra already feels borderline AGI but not quite there. I think as long as we keep getting ahead of the threats, it might be ok. Big ones are bio weapons probably.

KernelKind @KernelKind 1789014749

astra is definitely not borderline agi. definitely better than sol, definitely very good. but it still has way more to go.

i had it explain something to me and it was hallucinating a ton until i got pissed off & told it to seriously re-evaluate itself. on Ultra...