“I am personally accelerating the AI that will make me unemployed.
” The engineer who wrote DeepSeek v4.1's core attention operator acknowledges that AI will surpass his code in the next six months to a year, but says he will continue to optimize it. If I could be replaced, that hand would be mine. Official website
It seems that the strategy will need to be adjusted later
It seems that the strategy will need to be adjusted later: when the remaining 20% of the codex is left, GPT-6 Astra will be considered for strategy and execution, and the work will be left to DeepSeek V4.1, GLM or Hy4 You can even use the GPT-6 Pro on the ChatGPT web page to participate, and there is also Deep research on the web page. The web version is more fun to do research.
The cheapest AI model is often the most expensive choice in practice.
We assume a flash, meaning cheap and fast, model can handle any generative task. I ran the same prompt (a sailboat navigating between islands) across four models on and that assumption
DeepSeek V4 FULL 1 Hour 50 min Course
Want the SOP? DM me. 💬
You have reached the end of the archive
All of deepseek