Hey, welcome back! If you’ve watched a few of my videos, you’ve probably noticed I’m usually in Claude Code. Even when I’m brainstorming or working on a document.
The name makes it sound like a place just for developers. But what I care about is the setup around Claude: a folder to work in, instructions it can come back to, and skills that save me from explaining the same process every time.
This week I’m opening up my everyday setup. I’ll show you the features I keep coming back to, what I use them for, and how you can try them yourself. Some are quick wins, like asking a side question while Claude keeps working. Others have changed how I work, like starting a project at my desk and picking it up from my phone.
Here are ten things you can borrow from my Claude setup.
You can read through them below, watch the desktop walkthrough, or open the feature guide and pick something to try.
This newsletter is completely free. If you enjoy it, forward it to 2 or 3 friends or co-workers who are figuring this stuff out too. Someone forward it to you? Get your own copy here.
OK, let’s get to it!
– Derrick
Worth a look
Two worth trying this week, and one I’d wait on.
Worth trying: Claude’s new sheets and whiteboards
Claude can now make editable spreadsheets and shared whiteboards right beside the conversation. Both launched in beta on paid plans this week. Sheets starts with basic spreadsheet features and can export to Excel or CSV. What interests me is being able to change the thing we’re making while the conversation stays beside it.
Worth trying: Make your own little app in Muse
Meta’s new spaces let you describe a planner, tracker, or game and keep it inside Muse. You can change it as you use it and put it in your sidebar or main tabs. The small app you need might be something you describe instead of something you search for. Spaces are available wherever Muse is; sharing them with other people is still being explored.
Try this: Make a holiday gift tracker for three people. Add a budget, then ask Muse to change the layout so the people you still need gifts for show up first.
Wait on: Google Tabletop, a place to plan together
Tabletop turns a group’s links, files, and ideas into things everyone can use together: a dinner poll, a trip itinerary, or an expense splitter. A dinner poll everyone can change is more interesting to me than another itinerary buried in a chat. There’s a catch: it’s a Google Labs experiment rolling out through a US waitlist for adults 18 and older.
Make Claude part of how you work
The features I get excited about are the ones that make it easier to work on something with Claude: give it a process, ask questions along the way, look at the result, and come back to it later.
That’s why I keep talking about the “harness.” It’s the tools, instructions, and workspace around the model. Here’s what that looks like in my everyday setup. These are favorite features, not ten things that launched this week.
1. Give Claude somewhere to put the work
I give a project a folder on my computer. Even when I’m brainstorming, Claude can save a scratchpad, make a spreadsheet, or build a little page for me to review. The work has somewhere to live beyond the conversation.
2. Turn a repeatable process into a skill
My AI research skill contains the process I want Claude to follow when researching content. I keep my reviewed skills in GitHub and bring them into Claude as a plugin. When I improve the instructions, I can keep that collection in sync.
3. Give different jobs their own chats
In the walkthrough, I build a simple feature-guide website. Builder makes it, Design Review checks the design, and Copy Review checks the words. Projects let me keep those conversations together while giving each one a clear job.
4. Ask the question that just popped into your head
I love /btw for this. While Claude was building, I asked what it was building the site in. I could get an explanation alongside the work without turning the building conversation into a detour.
5. See what’s filling the conversation
The context view breaks down what’s taking up room in a session, including tools and skills. If a conversation is getting crowded, I can look at what’s in there. That’s a different question from how much of my plan allowance I’ve used.
Know someone who’d try this?
Forward this issue to a friend who has a project they want to make with AI. If someone forwarded it to you, get your own free weekly copy here.
Your link to share: {{rp_refer_url}}
6. Find the chat doing the heavy lifting
The project usage view shows which threads are using the most tokens. In my demo, it was the website builder, which made sense. It’s useful to see where the usage is going instead of guessing.
7. Give your skills collection a checkup
Run /skill-doctor to inspect the skills you’ve installed, when you used them, and their token usage. If you keep telling Claude “make this a skill so we can do it again,” that collection can get messy pretty quickly!
8. Let one chat check another
In Design Review, I asked whether the copy matched the design choices. Claude read the copy review and reported what matched and what needed adjusting. I used to copy answers between chats to do that. Now I can refer to the other conversation directly.
9. Leave your feedback on the actual document
This is one of my favorites. Have Claude put your thinking into a document you can comment on. Highlight the bit you don’t understand, leave a question, and keep reading. Then tell Claude, “I’ve left some comments. Address and respond.”
It took a correction in the demo: Claude first made an HTML file when I wanted a document I could comment on. So I asked for the format I needed.
10. Take the same project to your phone
I start the work on my desktop, then use Remote Control to review and respond from my phone. The same project and chats are there. My computer stays awake, online, and running the session; the phone is my remote into it.
Something to try this week
Don’t try to change your whole setup today. Open the feature guide, pick one thing you haven’t used, and give it a job in a project you already have.
If you’re not sure where to start, document comments are an easy one. Try this after a brainstorming conversation:
Put what we’ve worked out into an editable document I can comment on. Include the decisions we’ve made and the questions still open. I’ll leave comments on anything I want to change or understand better. Wait for me to finish, then address each comment and reply with what you changed.
If you want to see where the controls live, the video walks through my desktop setup. Menus and availability can vary with your version.
Someone building something interesting
What if AI could figure out how your LEGO model moves?
The builder behind LDraw Mecha, anteloc, has agents read the files for a digital LEGO model, work out which parts are mechanisms, and plan an animation in Blender.
I’m interested in how it figures out what moves. It starts with the model’s parts and structure, then turns a plan into an animation you can inspect. One of the examples is a pneumatic crane. The full model was too complex to animate, so the builder narrowed the job to the crane itself.
One idea to borrow: Give AI the source material for the thing you’re working on, then ask it to work on one small part. A model file, a spreadsheet, or your project files can give it details a screenshot would miss.
Go deeper
In this week’s video, I show the Code tab, skills from GitHub, the three project chats, side questions, usage views, document comments, and the same project on my phone.
Save the interactive feature guide. It includes examples to copy and a checklist. Your checklist progress stays in the browser where you use it.
Which feature would make the biggest difference to how you use Claude? Or is there something in your setup I should try?
Hit reply and let me know. I read every response (and try my best to reply as well)!
– Derrick
Forwarded this by a friend? Get next week’s Things to Try With AI here.