Anthropic CEO Dario Amodei: "I was inside OpenAI when GPT learned to code. I watched it happen in real time Ilya said: 'the models just want to learn.' I listened and I became enlightened then I left with 7 people and built Claude. not because OpenAI was wrong -... because I thought we could do it better " "the models just want to learn, get the obstacles out of their way - they'll do it " same is true for you, stop overcomplicating, stop asking permission, stop waiting for the perfect moment Claude ran 700 experiments in 2 nights - found improvements its creator missed in 20 years - it didn't wait, it didn't plan - it just started the gap between people who use Claude and people who don't is no longer months. it's years. bookmark and watch ↓show more

只要故事讲得足够动听， 分赃不均都能包装得如此神圣啊！

"Don't get in the way" "Delete your IDE" "Pay more - your usage ran out for running a simple loop" "You are not eligible to use this model" "I'm just a dick rider here to sell a shitty model that does the job you should be doing yourself"

这个傻逼越看越恶心。

This guy is a greedy Bug eater.

People underestimate how much the edge now comes from speed of trying, not “knowing AI.” But 700 experiments only matter if you know what to test. Isn’t taste becoming the real moat now?

Anthropic approach seems really promising

Inspiring message, but no model replaces execution.

Talks a lot, says very little

That’s awesome—believing in the shift and starting Claude shows real vision. Congrats on riding the AI wave and building something meaningful. Here's to continued breakthroughs.

700 experiments in 2 nights finding what took 20 years to miss is kind of nuts.

the real AI advantage isn’t knowing more it’s running more experiments than everyone else.

the honest answer is that the most valuable insights often come from watching failures, not successes. and sometimes you just have to leave and build something new instead of trying to fix what's already broken

🚨 BREAKING: Starting July 1, a new California law mandates that streaming platforms must ensure advertisements do not exceed the volume level of the primary content being watched, finally addressing long-standing viewer complaints about jarring commercial audio.

这段有点像 AI 版创业神话了

And here I am asking all four of them to rearrange and excel using a model of data from another excel and all four of them suck. Lies

Claude can only learn what's provided to it.

Dario Amodei did not write this but 40,000 people will quote it like he did that's a separate and more interesting AI story

Yeah very wow!!!! But we can't use. So for what that's all? Ok dream.. Goodnight

700 experiments in 2 nights thats search not research