ex-Anthropic engineer:
127,393 次观看
33 条评论
评论来自 X (Twitter) 上的原帖

@Serantych
the hard problem isn't making agents act-it's making them know when they're wrong
4 个赞

@0xCodila
to build such system, you need a right workflow for a right workflow, you need a right advice (it's in video)
1 个赞

@Granite0x
ai company engineers are straight up geniuses i wanna become one of them
4 个赞

@0xCodila
huh, you can! but this man a real professor and physicist, so to get hired - you should be not like everyone
1 个赞

@iiiichigo_chan
hope one day other AI will be at the same power level as Claude
2 个赞

@carm1nee
no way google engineer use Claude to train google models 😂
2 个赞

@0xCodila
hahha, yees Anthropic r win the race

@xatacrypt
claude still seems to set the bar for agents
1 个赞

@pawzzard
physicist hand codes nanoGPT just to get the job 💀 bro was NOT playing around 💀
1 个赞

@0xWast3
I was just looking for how to create an agent
1 个赞

@WorkflowStack
a physicist who hand coded nanoGPT just to GET IN and now hes back saying claude is still the one full circle arc hits different 🔥
1 个赞

@kiira7094
hope one day he'll make Google AI the best
1 个赞

@MAXdeg0
Claude’s agent game is strong right now.
1 个赞

@SlopToSignal
the 'loops of verification' part is the real unlock honestly most agents i see skip that and just trust the first output, which is how you get confident nonsense in prod

@heyraven_io
90 minutes is a lot of yap send me the link and i'll gist it for u

@Chntika_
pindah ke google cuma ganti meja, meeting sama deadline tetep numpuk bgt

@thatboyweeks
Why does it snap to Chinese a third through?

@Santahat24
I think.. I think.. awesome podcast 👏

@ironmind26
Just sitting to watch, thanks!

@bateshkaaa
It's very interesting to hear from an ex-Anthropic engineer that even at Google, Claude is still the leader when it comes to agents.










