I Found Claude Code's Secret Feature

0:00Claude code just dropped a feature that I don't see anyone talking about it and in this video we're going to be deep diving what it actually means and how it changes your workflow as a builder. So it's going to change how you build with AI forever. I want to break down what exactly it is. So I was scrolling X the other day and I saw this post from Anthropic developer with 19,000 likes. So I knew it had gained some traction. And all the post says is did you know about forward slash BTW in Claude code. So I'm thinking what the hell is forward slash BTW? I use Claude code every day. I use it to build out my mission control for my agent work, for basically everything. How did I miss this silent feature that went public? So it turns

0:45out there's a completely hidden command within Claude code called forward slash by the way that lets you have side conversations without losing your main thread. So when I say it's hidden, I mean hidden. It hadn't been publicly talked about and it's not in the documentation, it's not in the help menu. The only reason why we know about it is it because Anthropic dev decided to tweet about it. So let me break down what it actually does and why it's such a big deal. So normally when you're working with Claude code, you have this main conversation thread, right? You're building something, you're asking it to write code, make some changes, maybe debug some stuff. But then you have a question. Maybe you want to ask about best practices or you want to understand why it shows a certain approach or you just want to brainstorm a different

1:31feature. Now I had used the Claude code extension inside of cursor along with the terminal in Claude code running inside it and I use, you know, kind of as the extension as a back and forth on like changes it's making in my code base. So I feel like this is something similar but it goes a little bit deeper than that. So what most people do is like they ask the question right in the main chat and that can lead to context rot and you using the context window and filling it up with simple questions that you might have and that really does matter when you're trying to be accurate in building and shipping features cuz when you lose context, it's going to forget things that you wanted to do or things that maybe make state make mistakes in future code projects. This

2:16could be a big deal and I don't think people realize this but right now Anthropic only has a 200,000 token context window. So if you're chatty and you're back and forth with your Opus 46 or Sonnet, you could fill up that window incredibly fast. So when people ask the main question in the main chat and you're going back and forth, that's kind of when everything can go wrong because now Claude thinks that tangent, that side question is part of your main context. It starts incorporating that random discussion into its understanding of what you're trying to accomplish. Your clean focus coding session just became a confused mess where Claude is trying to optimize for three different things at once. That can be confusing. I see this all the time when I'm working

3:02with people in Shipping School. They'll show me their Claude code conversation and it's a winding path of half finished features or random questions and Claude getting more and more confused about what the actual goal is. But forward slash BTW, by the way, fixes that completely. So when you type forward slash BTW into your terminal or into your chat before a message, you're essentially creating a parallel conversation. You can ask Claude code anything you want and get a full detailed answer and have a good back and forth discussion without interfering with the main conversation. And none of that affects your main build thread. So it's like having a private sidebar convo while your main project stays perfectly focused. Think about it like this. You

3:47know when you're in a meeting and someone has a side question, so they'll lean over and whisper to the next person and that's what forward slash BTW does. The main meeting keeps going but you can have that side conversation without derailing everything. And this is a huge point of emphasis because that context window fills up much faster than you would think. You think 200,000 context token window is a lot when it comes to working with Claude but it will fill up much faster than you think and you will know that the quality of the code will continue to degrade as you keep on having these side conversations in your main chat. So the reason why this is so important is that context switching is the silent killer of AI assistant coding and I learned this the hard way building

4:32Easy Flip and Mission Control is that you just want to get straight to the point and very be very tactical with the prompts that you're giving your LLM. So Mission Control is the next JS dashboard I built that automates my entire workflow. Every morning 6:00 a.m. my agent called the Rizza scrapes trends, generates scripts, creates thumbnails, writes social media posts and gets a lot of content going, a lot of research. So when I was building this, I was constantly asking Claude code questions that weren't directly about the code. Like I'd be working on a trend scraping function, I'd suddenly wonder about rate limiting best practices. So I was kind of all over the place but I'd ask Claude about like rate limiting and it would start incorporating rate limiting concerns into every piece of code it wrote even

5:17when it wasn't relevant. Or I'd ask about like the database schema decisions and suddenly Claude is restructuring my entire data model when all I wanted was to understand the trade-offs. This is why I use that Claude extension in addition to Claude in the terminal because I could keep things separate and that's how I went about this after I experienced, you know, some troubles within the terminal and building out Easy Flip and Magic Hand is keeping two separate conversations can help me understand what Claude is doing in the terminal and in the code base and any side questions. And that's what I did before learning about this feature. Comment down below if you knew about this feature by the way. So my conversations became these sprawling messes where Claude was trying to be helpful but was actually making everything more complicated. It thought

6:02that every tangent and everything that we did on the side was actually a new requirement when I was just trying to talk about understanding what it was doing. So if I had known about forward slash BTW back then when I built Easy Flip, I could have saved a lot of hours. I could have kept my main build thread laser focused on actually shipping the feature I was looking to build while having all those learning conversations on the side. So let me talk about what this looks like in practice. So let's say I'm building a new feature inside of my mission control and I want to like add a calendar view for my content pipeline. I start by asking Claude code to help me build this calendar component like a clean focus conversation but then I have a question about whether I should be using a third-party calendar library or build it

6:48from scratch. So instead of asking that in the main thread and confusing things, I type forward slash BTW and ask about library recommendations. Claude gives me a full breakdown of different options, the pros and cons of course and implementation complexity. We have this whole discussion about calendar libraries. There's so many you can implement from Calendly and Google Calendar. But when I go back to the main thread, Claude remembers exactly where we were which is so nice to keep our organized thoughts organized and not just everything in the main thread. It's still focused on building the specific calendar component and that's what I asked it to do with the specific requirements I outlined. Then maybe I have another side question because you know there will be many when building and maybe I have a question about like accessibility considerations for

7:34calendar interfaces. Forward slash BTW again. Full discussion about our label, keyboard navigation, screen reader support but the main thread stays nice and clean. That's the difference between building with intention and building with chaos. I definitely built with chaos when launching Easy Flip. I was everywhere. I wasn't even considerate of any kind of context rot. I was just like, why are things breaking all of a sudden? And six months later it's thing things have changed. So here's what really gets me excited about this and the future of building with AI is like the psychology of behind why this works is what's fascinating. When you ask random questions in your main thread, you're not just confusing Claude, you're confusing yourself. Your brain starts to try to optimize for all of these different concerns at once. Like you

8:21basically lose track of what you were actually trying to accomplish. But when you use forward slash BTW for the side questions that you have, you're maintaining what psychologists call task segmentation. You're keeping exploratory thinking separate from your execution thinking. Your main thread stays action oriented while your side thread stay learning oriented. Thing is really important that, you know, I said this in another video is to learn what your LLM is doing and why it's doing it. You don't need to understand all of it but being able to figure out what's going on behind the curtain does help you as a AI builder and a person who uses AI to build apps. I truly believe this is how humans naturally think when we're building stuff. I have our main focus

9:07but constantly have little side thoughts like you little curiosities, little what ifs and I first started building with Claude desktop on another monitor, chat GPT on another monitor and Claude in the terminal on this monitor here and I was always going back and forth and that just confused the heck out of me. So the problem with most of these AI tools is they force everything into a one linear conversation which doesn't match how our brains actually work and think about thought processes. Forward slash by the way or BTW gives you that natural thinking pattern back. Like let me talk about some real examples of how I'm actually using this with my agent work is I'll have this agent called Atlas that acts as my chief of staff for all the other AI agents. This all happens

9:54within my mission control dashboard. Atlas coordinates all my other agents, make sure the daily cron jobs run smoothly, and all fire off. By the way, BTW, they all broke last night. So, also, if you're building with open claw and any kind of automation, prepare for things to break. It's just a natural part of the process. So, it handles all like the error recovery, all this fires off, I get a security audit at the end of each night as well, and basically handles the all thing the whole thing, right? So, when I was debugging an issue with Atlas last week, I kept wanting to understand how different error handling patterns would affect the overall system, but I didn't want to change the debugging approach based on those theoretical discussions. So, I used /BTW to have a deep conversation about the error handling philosophy and like the

10:40circuit breaker triggers, all the retry logic, and a lot of that stuff that goes into the troubleshooting process. But, meanwhile, the main thread stayed focused on a specific bug I was trying to fix, and I got the immediate little problem basically fixed. But, what I also learned a ton about better patterns for the future development is I just need to use that feature more often. So, when I'm working on Inspected Deck and that agent's doing all the generation of the podcast outreach and the newsletter and the X post generation, and basically I'll be implementing a new feature, but I will like have questions about the social media strategy or content psychology or platform specific best practices. /BTW lets me explore all that stuff without turning my coding sessions

11:25into like a marketing seminar. The main thread stays technical and focused, and the side threads let me be a little bit more curious and exploratory. I think if you are building anything with AI, the number one thing you need to know is that you need to be a lifelong learner, because what you master today may be irrelevant in weeks to come. So, being able to have a nice side chat with your whatever LLM you're using, but specifically Claude, this is going to be super helpful for people that are just getting going so you can understand some of the concepts and principles of shipping features and how AI apps are built. The real interesting part of all this is I think Anthropic kept this feature quiet on purpose. If everyone knew about /BTW, people would just start using Claude

12:10code very differently. Instead of these long windowing conversations, you'd have to have a highly structured interaction pretty clean in these main threads and an organized side discussion. That would probably reduce overall token usage because conversations would be more efficient, which is not what you want when most of the API token cost is a big part of revenue. People wouldn't need to repeat context as much because the main threads would stay focused. But, it would also change how people perceive the value of the tool. Right now, a lot of people judge AI coding tools by how long and detailed the conversations can get. Look, I had 200 messages and 200 conversations with Claude and it built my entire app. But, /BTW makes you

12:55realize that the longer isn't exactly better. Focused and dialed prompts, segmented in your main chat from your side chat, is what really matters. So, my theory is maybe Anthropic wasn't ready for that shift in user behavior. Maybe they wanted people to discover this feature naturally as they became more sophisticated. But, now that it's out there and with over 19,000 stars on X, people are starting to talk about it, it went viral, and I think we're going to see a big change in how AI and how people approach AI assistant development. So, how do you exactly start using this today? First of you're not using Claude code, what are you waiting for? You need to get on that. You need to at least build something with Claude code. It's honestly become my primary coding environment. I use it inside of cursor on the free plan.

13:40That's just my preference. You can use VS code or WinSurf or straight in the terminal, whatever. So, especially for anything involving AI and automation, I do love using Claude code for most of the building, and you need to start thinking about your coding conversations differently. So, before you ask any question, ask yourself, is this directly related to what I'm trying to build right now, or is this just a side curiosity? If anything, open up Claude.desktop or Claude in the web browser or ChatGPT and keep those conversations separate. If it's a side curiosity, try using /BTW to keep your main thread focused on shipping. Use the side threads for learning, exploring, and understanding the deeper concepts of what's going on in your framework and in your backend. Don't be afraid to have

14:26multiple /BTW conversations going. You can ask about performance considerations in like one thread, or you can ask about design patterns in another thread. You can explore deployment strategies, whatever you want. Just use that as a side conversation. You can do this all while your main thread stays locked in on one specific feature that you're building. Here's the thing is like this pattern applies beyond just coding. I'm already thinking about how to use this approach in other areas of building. So, for example, like when I'm working on the school community and content, I could have like a main thread focused on creating a specific lesson, but with the BTW threads for exploring different teaching approaches or understanding a certain student psychology or brainstorming future course topics. I don't know. Or when it comes to building

15:11new features for our SaaS EasyFlip or other SaaS products, I could keep them in the main thread focused on implementation while using the /BTW for market research or competitive analysis or whatever that might come into my head. I feel like most of us builders have some type of ADHD, that is for sure. And there's no problem with that. It's just using these tools correctly as they're designed is going to make you a much better AI builder. And if you're looking to be around a community of other builders just like this, you're going to want to join a community. I have it linked down below. We have 42 members in just 3 days. I want you to get in before the price goes up. We now actually have four boot camps a week with a fifth one on the way. My whole goal is to have calls 24/7 to help you

15:57wherever you live and wherever you're at to help you build and ship your tool or whatever you're trying to build. Maybe you're just trying to build an app for yourself personally. So, I guess the key insight after all of this is that building and learning are different cognitive modes. Most tools force you to do both at the same time, which makes you pretty much less effective as a human. That new feature /BTW gives you like a separate way to do them cleanly. That's exactly what we deep dive inside the community, not just surface level. Here's a cool new feature. How you doing? See you later. But, it's really understanding how these tools change the way we think and work, because that's what separates the people who just play with AI tools from people who actually ship products with them. It's not about knowing every feature, because if you did that, you'd be chasing the shiny

16:42object system syndrome every day. It's about understanding how to use these tools to match the way your brain actually wants to work and what your vision is for your business. If you want to learn how to build and ship AI apps the right way, join us down below. I'll put a link. We do four live calls every single week where I work with you hands-on to get your ideas built and shipped. You get access to the Next.js starter kit, and you actually get to learn and use these tools like Open Claw and Claude code effectively. Most importantly, you get direct help from me turning your ideas into real products. But, whether you join Shipping School or not, it's okay. Start using the feature right away. /BTU BTW in your terminal. It's going to change how you think about AI assisted development. It'll keep your

17:28main threads focused on shipping, and you could use your side threads for learning and understanding deeper concepts. You'll be able to watch how much faster you, you know, get items shipped and get stuff built, and it's just going to be a much cleaner way to build. If you guys haven't subscribed already and you like AI news and you like following the journey, then make sure you subscribe, and we'll see you in the next one. Be blessed.