Nvidia ceo jensen huang has never put these four numbers in a keynote
220 watts, 0.61 kwh a day, $5.12 a month, $200 once. the $200 is a used 3090 off ebay. that is a full replacement for a $400/mo opus 5 and gpt-5.6 stack, and it has been true since march the run above is that
Nartex's vision of AI - There is a lot of competition and that complicates everything.
Chinese models are getting better and better, several of them like Deepseek, KIMI, etc. They are already capable of competing with the most advanced American models, such as ChatGPT – 5.6 or Fable 5, many times,
Someone just rebuilt claude code in pure python and made it 230x cheaper to run
It's called clawcodex. a full from-scratch port of the claude code agent loop. 230k+ lines of python, MIT licensed it keeps your request prefix byte-stable, so DeepSeek's prompt cache covers your
This broke my brain a little
Moon dev: "deepseek is blowing me right now, oh my god dude, i'm ripping limits like crazy" "a lot of people come to me asking questions and they haven't even tried a single thing yet"
You have reached the end of the archive
All of deepseek