If you have a Claude subscription (and given the brutal limits on the free plan, you probably have one), you've likely used Claude Code. You don't need to be a developer or even "technical" to use it. As long as you can articulate your idea, spend some time learning the basics, and are willing to spend some tokens, you can go from idea to a full-fledged prototype within seconds.

Given that Claude Code's been all over lately, you've probably noticed that everyone has tips and tricks to share. There's no shortage of people telling you how to use it better (and we at XDA are no strangers to that game either). So, what could one more article possibly add? Well, there's one source that stands apart from the rest: Boris Cherny, the creator of Claude Code. He's previously shared his setup, which I've been using religiously since I first came across it. He's now shared a bunch more tips, and they all made my Claude Code experience noticeably better.

Use the /btw command for quick questions

Ask Claude a question without breaking its focus

The entire premise of Claude Code is interacting with it using completely natural language. You describe something in plain English, and it figures out the rest. Pretty simple, right? However, there's a layer beneath that most people don't really touch: slash commands. These are simply shortcuts you type starting with / to control various behaviors. For instance, you can type /model to switch models, /clear to reset your conversation, /cost to see token usage statistics, and more. You can simply type / to see every command available to you.

The /btw command is something Claude Code introduced in March 2026, and Cherny shared that he uses it all the time. The command stands for "by the way," and it does exactly what the name suggests — lets you ask a quick side question while Claude is mid-task without interrupting its work or cluttering your conversation history. The answer pops up in an overlay, you read it, and Claude keeps going like nothing ever happened. It doesn't interfere with what you already have going.

When the /btw command was first announced, my first thought was, "Why not just open a new Claude Code session in parallel?" In fact, that was one of Cherny's setup tips: running multiple sessions in parallel. The best part about this command is that side questions have full visibility into the current conversation, so you can reference anything Claude has already seen or done in that session. A new parallel session starts with a blank slate, whereas with the /btw command, you can ask things like "what was that file you looked at earlier?" and get an answer.

The Claude in Chrome extension lets Claude see what it builds

Give Claude a pair of eyes

If you've been using Claude to vibe-code different tools, there have surely been times where it's given you the output, you go ahead to test it, and it just doesn't...work. Then you're stuck going back and forth, describing what's wrong, hoping Claude gets it right on the next try. A lot of the time, Claude will ask you to do something while running the tool on your browser and tell it the output, perhaps check the console for errors, describe what the page looks like, or tell it what happens when you click a button.

The problem, as Cherny describes it, is simple: Claude can't see what it built. Instead, you essentially become Claude's eyes. This is exactly what the Claude in Chrome extension fixes, and you're missing out if you haven't tried it yet. This extension is available in beta to users on any Claude paid plan, and support documents explain that the extension and Claude Code "work together for a build-test-verify workflow." This browser extension basically gives your Claude Code session a browser.

It gives it the ability to see, interact with, and test what it builds in real time. For instance, I recently used Claude Code to create an extension that would let me see if someone's seen my Instagram story. I wanted to add a search bar that would let me search through the list of viewers. I was stuck in a loop of testing the extension, explaining what wasn't working to Claude, only for the next attempt to break something else. I then turned on the Chrome extension, and Claude spotted the problem, fixed it, and moved on!

As Cherny puts it, you wouldn't ask an engineer to build a website without letting them use a browser. So, the most important thing you can do with Claude Code is give it a way to verify its own output. Once it can, it'll keep iterating until the result is right.

The /loop and /schedule commands for hands-off workflows

Set it, forget it, let Claude handle it

Even if Claude isn't the AI tool you rely on daily, you likely have the habit of running the same prompts over and over again. You run these prompts manually, get a result, and then type the same thing again 10 minutes later. This is something that requires a lot of babysitting and back-and-forth, and it's certainly something that should be automated. That's where two more slash commands come in: /loop and /schedule.

The /loop command lets you set a prompt to run automatically at a set interval. You give it a time and a task, and then Claude re-runs the prompt on the schedule for as long as your session is active. For instance, say you have Claude connected to Slack and your Gmail, you could set up a loop to monitor both every 30 minutes and surface anything that needs your attention. You can also omit the interval entirely, and Claude will figure out the timing on its own! As mentioned above, the /loop command only works while the session is open. Close the terminal and it's completely gone!

That's where the /schedule command picks up. This lets you set up persistent tasks that run even when you're not actively in a session. There are two types: Desktop tasks and Cloud tasks. The former run on your machine as long as the app is open, and the latter run on Anthropic's servers even when your device is turned off.

Cherny runs multiple loops in parallel! This includes things like automatically addressing code review comments, putting up PRs based on Slack feedback, and closing out stale PRs. He recommends experimenting with turning your workflows into skills and loops.

The --add-dir flag for working across folders

Give Claude the full picture

The first thing Claude Code does is ask if you'd like to grant it permission to access your current working directory. That's the folder you launch it from, and by default, that's all it can see. It might ask you to grant access to additional folders along the way, but you can save yourself the hassle by using the --add-dir flag to give Claude access to additional folders upfront.

For instance, say I wanted to give the tool access to another project while working on the Instagram Chrome extension for inspiration or to reuse some code. I could launch Claude with claude --add-dir ~/Projects/old-extension, and it would have access to both projects right away! Cherny mentions that if you're always working across the same folders, you can also add them to your settings.json so they load automatically every time.

Use Claude Code from your phone, browser, or terminal

Who says you need a desk to code?

Three of the tips Cherny shared on his X thread all point to the same idea: Claude Code isn't chained to your terminal. The tool is far more portable than most people think, and you can use Claude Code from pretty much any device.

First up, something a lot of people don't realize is that Claude Code has a mobile app too. Cherny shared that he writes a lot of his code from the mobile app, and that it's a convenient way to make changes without needing to open his laptop. You can open the Claude app on iOS or Android, head to the Code tab, and begin!

Second, you can move sessions between devices using the /teleport command. You can begin a cloud session, and then use it to pull the session into your local machine and keep building! You can also use the /remote-control command to control a local session from your phone or browser.

Finally, Dispatch is a feature I've been using for a while now! It turns your phone into a remote control for the Claude Desktop app, and lets you do pretty much anything on your laptop without being anywhere near it. Cherny says he uses Dispatch daily to catch up on Slack, manage files, and handle email. He says that when he's not coding, he's dispatching.

The creator's tips hit different

There's simply no one else you can trust than Boris Cherny himself. Every tip he's shared so far has made a noticeable difference in how I use Claude Code, and I'd bet the next ones will too.