Video yükleniyor...
Video Yüklenemedi
Creator of Claude Code just dropped a 6-min workshop on new Claude feature during live session in London. Boris Cherny: “A lot of my code these days is written by "routines". I’m not doing the prompting - I create the routines that do the prompting.” 6 minutes. Free. From... show more
1,688,489 görüntüleme • 4 ay önce •via X (Twitter)
42 Yorum
That’s an insane feature movez!
Yeah, Ruj. Just saw it at the London live session. This is crazy.
Crazy man!
This guy never misses. Useful as always.
Watching Boris’s live session feels like watching how devs will build in the coming months.
Routines only become leverage when they have a contract: trigger, allowed files/tools, stop condition, and the review packet they leave behind. Otherwise it’s prompts with a nicer name.
Have you tried creating such routines after this workshop? How much have they changed the way you work?
they just dropped it on live session, im on my way to test it bro
Cool, bro) Are you on your way to testing? Then post how it went and if it really changed your routine.
Routines writing prompts that write prompts is where leverage actually compounds. But Boris skips the hard part: when two routines conflict at 3 AM with no human in the loop. Constraint resolution between automated workflows is genuinely unsolved. That is the real frontier.
I think that is the problem. The tool was originally built for internal users that get free/unlimited tokens. Also code was not added to tell the tool to ask the consumer end users all these things when setting up. OpenClaw knows how todo that.
the real unlock isn't just writing routines, it's figuring out which ones to use and when they break, so you don't end up with a product that's 10x more complicated than it needs to be
conferences like these help us keep up with the breakneck pace of development
yeah, only the cutting edge of development alpha here
The trader version of this is writing strategies that deploy strategies Same idea, different surface
already hear this performance
@0xMovez kinda wild how short sessions can pack a punch. routines doing the heavy lifting now huh?
this is really useful man
Just saw it at a live session in London, and it looks really interesting to me.
it’s a fundamental shift in AI development
yeah, if you know how to use it correct, it will change the way you code forever. btw are you using Claude in your daily routine ?
It can be said that Claude and Anthropic are part of my digital DNA😁
6 minutes of pure gold. Boris is spot on—the real leverage is building systems that prompt for you.
yeah true, man jump in Claude Youtube chanel and watch full live. This will change your coding routine
we're one step away from humans just being observers
Everyone must know about this amazing sessions.
You can't miss this, I'm going to watch
the shift from manual prompting to orchestrating routines feels like moving from assembly to C
Same shift on my side. Stopped writing prompts, started writing skills. The skill outlives the chat.
Most people using 10% of what Claude can do is painfully accurate. The setup most people skip is real 👀
automating the actual prompting is peak meta engineering
This could really streamline coding workflows!
新機能のワークショップのライブセッションで6分のクロード機能の新しい紹介。感動的。
Erm, when you run something on a schedule it doesn't make it higher order anything. 😂
I am absolutely mind blown by his take on routines
They really love those sweaters at Anthropic
this is the actual shift. you stop writing prompts and start composing routines. the routine is the unit of work, the model is the renderer.
That live session was gold. The part where he debugged a complex refactor live in under 3 minutes was genuinely impressive. Workshops > blog posts for learning agent workflows.
The people getting the most out of AI are slowly moving beyond prompting. They are building systems, workflows, and reusable routines that think in layers. That is the real productivity unlock. Not one perfect prompt.
from prompts to routines is the next leverage layer
Spotify hit 4,500 deploys/day partly because of this shift, engineers stopped writing code and started writing the systems that write code.
by virtue of the "owner" of the LLM having an unlimited budget on the most expensive model makes anything new a shill job. noise unless i see it in the wild
