These 8 Simple Steps Will Change How You Use Claude Forever

Step #1:

Most people start every Claude conversation from scratch - that's a waste.

Claude has a built-in memory feature that lets you control exactly what it holds onto between sessions.

The crazy thing is, 99% of people just let Claude build passive memories about them, and they never actually edit/refine what they want Claude to know.

In Settings → Capabilities, you’ll find “Memory.”

We highly recommend taking a few seconds to edit your memory from chats and even to import memory from other AIs (like ChatGPT).

Tip: You can have Claude build this context for you - just prompt a new chat: “I’m optimizing my Claude memory, ask me 25 questions that will help build an accurate summary.”

Step #2:

Claude writes in Claude's voice by default. Which is fine, but oftentimes sounds heavily like AI.

If you want to eliminate “AI slop” in your Claude text outputs, this step is key.

In a new chat, you’ll find “+” where you can scroll down and select “Use Style.”

If you’ve never touched this before, it will default to “Normal.”

By clicking “Create & edit styles,” you’ll be able to give Claude a custom voice.

You have two options:

Complete game changer.

Step #3:

Memory tells Claude who you are for general chats, but custom instructions tell Claude how to behave inside specific workflows.

To make it easy, just think of custom instructions as a completely separate context from the memory context we set up in Step one.

Go ahead and create a new Project, and on the right-hand side menu bar, click “Instructions.”

This is where you can set the complete context for whatever project you're working on (investment advice, career consulting, etc.)

Simple, yet highly effective tweak you can make to get exactly what you need from Claude outputs.

Step #4:

This is an underrated tip that we don’t see people using enough.

You can build custom Slash commands directly into your Claude agent.

Example: you create a “/grammar” command that triggers Claude to conduct a grammar check on the following text.

Yes, you can use Claude Skills for this, but this strategy is also a good tool for quick & easy instruction sets.

The possibilities with slash commands are endless - you can set them up for any prompt you find yourself having to retype.

Step #5:

Piggybacking off of the last step, you can also build “Skills” that help automate repetitive workflows.

Skills are essentially instruction files for Claude to follow.

Example: “Use my grammar checker skill to optimize this text [x].”

Skills is our personal favorite Claude workflow, and every team member uses at least one.

In Customize → Skills, you can build them.

Tip: Claude has an entire Skills Directory - if you’ve never built a Skill before, this is a great starting point.

We recommend building slash commands for simple workflows/prompts, and taking the time to build Skills for more complex workflows.

Step #6:

If you do only one thing from this entire list today, make it this.

Connectors take just 30 seconds to set up, but they let Claude become 10x more productive and accurate by accessing your tools.

Example: You connect Claude to your Gmail, which allows Claude to read your entire inbox.

Another Example: You connect Claude to your Google Calendar, which allows Claude to send invites, read your schedule, and more.

You’ll find the option to add connectors from Customize → Connectors.

Our essentials: Notion, Gmail, Google Calendar, Excalidraw, Slack, Figma, Google Drive.

Step #7:

Claude Skills are level one for AI automation.

Setting up Cowork Plug-ins is the next step.

While Claude Skills can only automate one workflow, Plug-ins are built to automate entire “roles.”

They package multiple skills, multiple connectors, and instructions into one central hub.

What to do:

Step #8:

This final step is arguably the most game-changing.

Most people have no clue this exists.

Claude in Chrome is an official Anthropic Chrome extension that lets Claude live in your browser.

It can organize tabs, view your current tab, and more.

It’s a very easy and efficient way to access Claude quickly and share your desktop.

You can download it here: https://chromewebstore.google.com/publisher/anthropic/u308d63ea0533efcf7ba778ad42da7390

So, there you have it, the exact 8-step roadmap to customizing your Claude to unlock productivity that most people don’t know about.

Now, let’s dive into the Midweek Edge - our manually curated list of the top AI updates from the past week.

In last week’s newsletter, we speculated over Claude “Mythos” - Anthropic’s most powerful LLM ever built, specifically for cybersecurity.

Last night, Anthropic confirmed and officially released Mythos in preview.

The crazy part is, they said it’s “too powerful” for the public.

Read Anthropic’s official preview here:

Google just released its most powerful family of open-sourced LLMs.

Gemma 4 is available now, and 100% open-sourced, with the ability to run offline.

Read more here:

In an effort to improve cybersecurity in the AI era, Anthropic has officially released “Project Glasswing” - a cybersecurity project involving tech giants Apple, Anthropic, NVIDIA, Amazon, Microsoft, and more.

Read about Project Glasswing here:

Looking Ahead - The top AI trends we’re closely monitoring