NEW Claude Code Upgrade: /btw + /voice

Anthropic just quietly released two new features for Cloud Code that are going to 10x your productivity. And for one of them, currently only 5% of people have access to it. So the first feature that

was announced is the /BTW feature or by the way feature for quad code. It was

announced by Tariq who's on the technical team at Anthropic. He made

this post here on X. And so what the heck is it? One of the main problems that I've had when I'm coding with quad code is that you can't really interrupt its process in like an effective way in a way that's not going to cost you a

whole bunch of tokens. Well, the

slashBTW feature totally fixes that. So,

for quick questions that don't need to stay in context, you can use the slashbtw feature. The answer is going to

slashbtw feature. The answer is going to appear in like a little dismissible overlay. It's never going to actually

overlay. It's never going to actually enter the conversation history. So, you

can kind of like interact with cloud code, ask it questions, have a conversation with it without actually interrupting its workflow or without costing you a whole bunch of tokens.

This is a lot easier to demonstrate. So,

I'm just going to open up a terminal window here. I'm going to pop into cloud

window here. I'm going to pop into cloud and let me just show you what this thing does. So, I'm going to say I'm going to

does. So, I'm going to say I'm going to create a process that's going to get trending news from Reddit and write a LinkedIn post for me. And right now, the way that I typically interact with this, like if I have another thought, like I

can type this into the chat right here and say, "Actually, never mind. I don't

want to do Reddit. What if I did X instead?" I can send this off, but

instead?" I can send this off, but what's going to happen is I can't actually have a back and forth with Claude code about the changes that I want to make without it costing tokens, without it actually getting involved in

the process or starting to go to work on the task. What the BTW feature allows me

the task. What the BTW feature allows me to do is let me just go ahead and start a new conversation. So, what the BTW feature allows me to do is I can start the same task about processing news from

Reddit and writing a LinkedIn post for me. But now I can type in slashbtw and I

me. But now I can type in slashbtw and I can say, do you think it would actually make sense for me to grab ideas from X?

And what about creating a tone of voice guidelines? And so now what I can do is

guidelines? And so now what I can do is I can actually keep Claude getting to work on the original task itself while still having kind of a side conversation with it. Now, previously what I could do

with it. Now, previously what I could do is like I could open up a new terminal window and spin up another version of Claude or open up another agent to have this conversation. But the problem is

this conversation. But the problem is that this agent doesn't have any context from the existing conversation. And so

it's not really going to know what we're talking about. And so in reality, like

talking about. And so in reality, like this isn't super helpful for us to continue our planning. And so we can see that the BTW response comes back with this little yellow tag here. It's in

this side panel and just says yes to X.

Definitely. Reddit is great for raw pain points and discussions, but Twitter is where the actual AI and tech discourse happens in real time. So, it's kind of allowing me to have this conversation with it. And what you can do is you can

with it. And what you can do is you can press spacebar, enter, or escape just to get rid of this workflow. Or you can continue to chat with Claude while it goes to work on your process. One of the limitations of this is that Cloud Code

doesn't actually keep a memory of your conversations in the BTW command. So, if

I wanted to implement something like this, what I would say to do is you probably just have to go ahead and copy this. And now if I went ahead and

this. And now if I went ahead and pressed space to exit out of that conversation, I would say like I want to create some tone of voice guidelines.

Here's some information about that. And

then I can go ahead paste in the response that Claude Code gave me before about my core origin story, my specific offers, my audience, and things like that. And so now Claude can get back to

that. And so now Claude can get back to work. I can go ahead and instruct it to

work. I can go ahead and instruct it to make changes. I can have another side

make changes. I can have another side discussion with it if I'm so inclined.

So in a nutshell, the BTW command is a way for you to get quick answers without polluting the context at all. So before

this command, what you would do is you would ask a question of cloud code and that question and the response would get injected into the conversation history.

So it would cause a lot of bloat. It

would increase the context window. You'd

either have to compact or clear things more frequently. Your context is going

more frequently. Your context is going to get cluttered, especially if it's something that you're like unsure of like the question I asked about like using X over Reddit. Like I didn't want that to actually be a part of the conversation. It's more of just an

conversation. It's more of just an aside. So your performance is going to

aside. So your performance is going to degrade over time without having these kind of separate talks. So now if you use the /BTW command, the answers are going to appear in that sort of overlay and when you dismiss it, it vanishes

completely from the conversation and from the history and so your context stays clean. And so in some cases you

stays clean. And so in some cases you might want this. Just keep in mind it's not going to have access to the conversation that you had with it. A

cool way to think of this is BTW is basically the opposite or inverse of what a sub agent does. So if you use BTW, it'll see the entire session. So

your entire conversation history, anything that is already built, but it won't have the ability to call any tools or take any sort of action for you.

Whereas a sub agent won't see the entire session history or the context of everything, but it is able to have access to tools and actually take action and build things for you. If you don't have access to this, just make sure you're running the latest version of

Claude Code. All you have to do is type

Claude Code. All you have to do is type in cloud and update and it's super super easy. This will go ahead and it'll check

easy. This will go ahead and it'll check for the latest update and it'll update everything for you. Likewise, if you're inside of some sort of like vibe coding platform like anti-gravity or VS Code, just come over to your extensions. You

can go ahead click into the cloud code extension over here. Come into this little arrow and you can install a specific version and just make sure that you're installing the latest version of this. And for me, I always just have

this. And for me, I always just have auto update on to make sure that I'm always up to date. Now, the second feature that they announced, I know a lot of you are going to like cuz I get comments on my videos all the time that are like, "Yo, like how are you talking to your machine? Like, what are you

using?" I use an app currently called

using?" I use an app currently called Whisper Flow. So, I just press the

Whisper Flow. So, I just press the little function version on my keyboard and basically I can talk to any single application that I have. But what's cool is that Claude Code is actually coming out with voice mode. Right now, it's

only live for 5% of users. But over the next coming weeks, they're going to be ramping this out to everyone. So,

basically, all you have to do is type in /voice and you'll be able to talk to QuadCode just like I am now, except you're not going to have to pay for a third party service to do it. So now you have everything you need to 10x your

productivity and reduce token cost with the slashbtw command so you can have conversations in parallel and pretty soon everybody should be able to talk to cloud code directly using their voice.

Let me know in the comments which of these features you think is cooler.