← Back to all free guides

New To Claude? Do These
5 Things First, Part 3

Five settings hiding in your account that most people never touch. Where to find each one, and the exact move that makes it worth having.

Part one made Claude know you. Part two made it do real work. This set is the stuff most people never turn on, then wonder why Claude feels like a chatbot from 2022.

All five already exist inside your account. Memory and Claude Design work on the Free plan. Cowork and Research need a paid plan. Each one gets where to click and, more importantly, what to type once you’re there.

Haven’t done the first two sets? Start with part one and part two.

The 5 Things

01

Turn on memory, then go read it

Claude writes down what it learns about you and carries it into new chats. The step almost nobody takes is opening it up and reading it.

Where: Settings, then Your Claude. Make sure Memory from chats is on (it’s on by default for individual plans), then click Manage next to Your memory. It’s sorted into topics you can open, edit or delete.

The move: read it like someone else wrote it about you, because something did. A wrong memory is worse than none, because Claude builds on it for months. Mine still thought I was working on a project I’d dropped in March. Run this to clean yours up.

Memory Audit
Show me everything you currently remember about me, grouped into: my work, my preferences, my ongoing projects, and facts about people in my life. For each item, tell me where you picked it up if you know. Then flag: - anything you're unsure is still true - anything that contradicts something else you have - anything that looks out of date Don't soften it and don't fill gaps with guesses. When you're done, ask me about the flagged items one at a time, and update your memory with my answers. Finish by listing the 3 things you'd most like to know about me that would make your answers better.

Anthropic’s help page →

Want the deeper version? Read the Claude memory guide.

02

Open Claude Design

Describe a landing page, slide deck, one-pager or graphic, and Claude builds it on a live canvas you can edit directly or redirect in chat, so you never start over.

Where: the Artifacts tab has Design templates to start from. In a chat, you can also pick Output, then Design, in the message box. It works on every plan, Free included.

The move: set up your brand before you make anything. On paid plans you can save a reusable design system from your logo, fonts and brand files. On Free, paste this at the start. It’s the difference between a generic page and one that looks like yours.

Brand First
Before we make anything, I'm setting up my brand so everything you build stays on it. - Display font: [FONT]. Body font: [FONT]. - Colors: [HEX] main, [HEX] background, [HEX] text, [HEX] accent. - Logo: [attached / describe it]. - My brand sounds: [three words, e.g. warm, direct, no fluff]. - Who it's for: [my audience in one sentence]. Rules: never use a font or color I didn't list, no stock-photo style imagery, never center long paragraphs, and keep text large enough to read on a phone. Confirm what you've got back to me as a short style sheet, tell me anything that's missing, then ask me what we're making first.

Anthropic’s help page →

Full walkthrough in the Claude Design guide.

03

Hand off work from your phone with Cowork

Cowork is Claude doing a multi-step task for you instead of just chatting. On Pro, Max and Ultra, new Cowork sessions run in the cloud, so you can start one from your phone and it keeps working with your laptop closed.

Where: the message box, on the web or in the app. Select Cowork and describe the task. You can check its approach and steer it while it works. Anything that touches local files on your computer needs the desktop app open.

The move: use it for the thing you always remember at the worst time, like standing in line. Send the task and it’s done when you sit down. Start with a read-only job like this one.

First Cowork Handoff
Go through [my Google Drive folder called FOLDER NAME / my inbox from the last 7 days] and tell me what's actually in there. 1. Group everything by what it's for. 2. Name anything that looks like a duplicate, an old version, or something I forgot to reply to. 3. Tell me the 3 things I most likely need to deal with this week, with one line on why. Rules: don't move, rename, delete, send or reply to anything. Just report back in a short list I can read on my phone. If you can't reach something, tell me instead of guessing.

Anthropic’s help page →

More in the Cowork guide. Dispatch, the older phone-pairing feature, is closed to new users, so Cowork is the way in now.

04

Change your effort level

Effort tells Claude how hard to think before it answers. Most people never touch it and burn their limits on tasks that didn’t need it.

Where: click the model name next to the send button, then Effort. The levels are Low, Medium, High, Extra high and Max, and the one marked Default is the model’s recommended setting. You can change it mid-chat, starting with the next reply.

The move: treat it like a budget. Use Low for reformatting, renaming and short summaries. Save Max for the few things a week where being wrong costs you, like a pricing call or a hard email. Higher effort is slower and uses your limit faster.

The Weekly Sort
Here are the tasks I bring to you in a normal week: [LIST 8 TO 10 REAL TASKS]. Sort every one into LOW, HIGH or MAX effort. - LOW: the answer is already decided and you're just producing it (reformatting, short summaries, renaming). - HIGH: real reasoning, but a wrong answer is cheap to fix. - MAX: being wrong costs me money, time or credibility. For each task, give me one sentence on why it landed in that bucket. Be strict. Most tasks belong in LOW. Finish with a one-line rule of thumb I can remember.

Anthropic’s help page →

05

Test Research

Instead of one search and one answer, Claude runs search after search, each one guided by what the last one found, across the web and your connected apps.

Where: click + at the bottom left of the chat and pick Research. Web search has to be on. Start it on web or desktop if you want it to search your connected apps too, because on mobile it only searches the web.

The move: give it a decision, not a topic. "Research email platforms" gets you a table you could’ve found yourself. Name your constraints and ask for a pick, and you get an actual answer.

The Decision Brief
I need to decide [THE ACTUAL DECISION]. My constraints: - Budget: [X] - Timeline: [X] - My technical level: [X] - It has to work with: [THE TOOLS I ALREADY PAY FOR] Research the real options, then give me: 1. One recommendation. 2. The two runners-up and exactly why they lost. 3. What I give up by taking your pick. 4. The one thing that would change your answer if it were different. Use current sources and cite them. Tell me where the evidence was thin instead of writing around it. Keep the final answer under 400 words, with the sources at the end.

Anthropic’s help page →

Take All Five With You

If You Only Do One

Do the memory audit. It makes every future chat better without you doing anything else, and it’s where people find something wrong sitting in there.