talked with a programmer today, he is working as a data analyst
"we use chatGPT at work, it writes a lot of code"
"oh, do you use Codex?"
"what is codex? we use Claude"
"that is not the same as chatGPT"
"it is the same to me, I call them all chatGPT"
"ok, do you use claude coworker then?"
"what is coworker? no, I said we use claude, you have no idea haha"
"do you have like a company plan, or?"
"we all use the free version, it is good enough"
ok, I am out of here... they will be losing their job soon because they refuse to tokenmax
we are in a bubble
はい、ローカルLLMはレートリミット無しでCodexに迫るかのような性能で無限に使えます、Mac Studio 512GB構成を2台用意してThunderbolt 5によるBridge接続または10Gb Ethernetでノード間通信する分散推論環境を構築し、1Tパラメータ級MoEモデル Kimi K2.6を���行すればいい感じで使える理解でいいです