Considering everyone is talking Claude Code recently (even people who once rejected AI coding tools altogether), your feed is probably full of posts about the ideal setup, secret tips and tricks, and underrated features you’re missing out on. We at XDA are no exception. We absolutely love it Claudeand we’re always writing about ways to get the most out of your subscription.
But at the end of the day, we can only tell you so much. We can learn what the community is doing by going to X, Reddit, and other forums (which we do anyway), but the real golden advice comes directly from the people who built it. Claude Code was created as a side project by Boris Cherny, who always shares his setup and tips on X. Anthropic has released Opus 4.7 Cherny a few weeks ago and with the launch threw the rope shares his top tips for getting the most out of the new model.
Want to stay up-to-date on the latest developments in artificial intelligence? The XDA AI Insider newsletter drops every week with deep dives, tool tips, and hands-on coverage you won’t find anywhere else on the site. subscribe by changing your newsletter preferences!
It suggests to enable Auto mode
Stop Claude from looking at the baby with each click
Claude Code prompts you to move forward by pausing in the middle of a task and waiting for you to confirm the action before continuing. Every file you edit, every terminal command you want to run, every dependency you want to install, every time you want to search the web, etc. It’s a smart security measure and it’s there to make sure you’re in control and nothing happens without your say. This can seriously slow you down when you’re locked out.
Anthropic was well aware of this, so it enabled permission modes that controlled how often Claude paused to ask. There are several to choose from: Default, Accept it Edits, Plan, Don’t askand Skip permissions. Each mode pre-confirms certain actions and commands. It’s the newest addition to the lineup Automatic regime, and this is the regime that Cherny swears by. In an X post, Cherny shared that Opus 4.7 “loves to handle complex, long-running tasks.” He shared that this is iterative until a performance benchmark is reached, and while that’s great when you’re building, it also means you’ll have to do dozens of manual thumbs ups throughout the process.
you can use Pass by Permissions mode here, but the word “dangerously” in its official name for a reason. Automatic mode is the sweet spot. It allows Clod to run without constant permission requests (which is the advantage you get with Bypass Permissions), but model-based classification decides whether each action is safe or not before it’s done. Routine things like local file edits and dependency installs go without a hitch, everything else is really blocked and risky.
Cherny explains that this also means you can run more Claude Code sessions in parallel (which was one of his jobs). tips on its previous installation). Unfortunately, Auto mode is only available on the Max, Team, and Enterprise plans. So if you’re on the Pro level, you’ll have to stick to one of the other permission modes.
Use /az-permission-prompts to clear your permission list
You have received permissions for nothing
There are probably some permissions that you approve on autopilot without even thinking about it. However, in spite of you always Claude Code who approves them doesn’t really know and will keep asking you every time. Well, Anthropic has quietly rolled out a /fewer permission requests skill that fixes this, and here’s the next tip in Cherny’s thread.
This skill scans your session history and detects the secure commands you have confirmed. Finds both bash and MCP commands. It then gives you a list of commands it recommends you add to your permission list. One run and you’ve cut out a lot of hints you never needed in the first place!
Use repetitions to pick up where you left off
helps you get through wait where was i moment
If you’re like me, you run multiple sessions of Claude Code and are constantly switching between them. The problem is that by the time you get back to one, you’ve completely forgotten where you left off. Anthropic has launched a summary feature that addresses exactly this. When you return to a session after you log out, Claude Code shows you a one-line summary of what’s happened in the session so far.
This summary only occurs when at least three minutes have passed since the last completed loop and the terminal is “unfocused”, meaning you’ve moved to another window or away from the terminal entirely. It won’t block you mid-stream. Instead, it prepares silently in the background and is ready the moment you switch back. While I saw the recap feature while using Claude Code a few days ago, Cherny’s thread is how I realized that you can also run the /recap slash command at any time if you want an on-demand summary.
Use /focus to see only what’s important
Just show me the results
With how fast Claude Code ships these days, there are always some features that manage to slip through the cracks. Focus mode is one of them and I didn’t even realize it was down until I saw Cherny’s thread!
This mode boils down your session to just three things: your request, a summary of the tools used by Claude, and the final response.
Files it reads, commands it runs, step-by-step reasoning, etc. as all intermediate things are hidden. You just see what you ask for and what you get back. You can turn the mode on and off using the /focus command.
Configure your effort level to think smarter
Tell Claude how hard it is to think (yes, really)
If you’ve been using Opus 4.7 lately, you may have noticed something called Adaptive Thinking. This mode basically allows Claude to judge how deeply to think about each step based on how complex the task is. Simple things get answers quickly, harder problems require more reasoning.
Although Adaptive Thinking automatically controls the depth of thinking, thankfully you can still influence the overall intensity through effort levels. Opus 4.7 offers five: low, medium, high, xhigh and max. Cherny says he uses xhigh for most tasks and only maxes it out for the hardest ones. Xhigh is now the default in Claude Code, and you can change it at any time using the /effort slash command.
He also recommends giving Chloe a way to check her work
Let Claude see what he has built
His final piece of advice is something he says “there’s always a way to double or triple what you’re getting from Claude,” which he explains is more important than ever with Opus 4.7. The tip is pretty simple: Give Clod a way to check his work. He explained that there are several ways to do this, and it differs depending on the position. For example, if you’re doing some frontend work, you can Use the Claude in Chrome extension Getting Clod to control your browser. If it’s a program, you can use Use My Computer to get Clod to test its output.
It’s something I’ve done a little bit myself and it’s changed the way I work. I vibe-coded a Chrome extension using Claude Code a few days ago and it just didn’t work. The extension required some manipulation of Instagram’s DOM, and no matter how many times I described the problem, Claude couldn’t get it right. As soon as I gave access to my browser via the Chrome extension, it was able to see what was wrong and fix it on its own. At the end of the day, there’s only so much you can explain in a text request. Sometimes Claude just needs to be seen for himself, and as I would like to describe, the Chromium extension gives Claude the eyes.
A few small changes go a long way than you might think
It’s easy to brush past tips like these and think they’re too simple to matter. But after spending a few days using Claude Code with the above tips in mind, I can confidently say that they make a difference. If the tips come directly from the person who built the thing, it’s probably worth a try.
