Every Claude habit ranked by two things: how much it pays off, and how long it takes to set up. Start in the gold corner: six moves, five minutes each, massive payoff.
5 Minutes + High Payoff (start here)
Turning Memory on · about-me.md · new chat per topic · asking Claude to push back · voice mode brain dumps · picking the right model per task
A Weekend + High Payoff (build next)
Claude Code · your first Skill · your first MCP connector · a Project per area of your life · connecting Gmail + Calendar
5 Minutes + Low Payoff (the illusion)
"Write me a..." one-liners · re-explaining yourself every chat · asking 3 AIs the same question · saving prompts you never use
A Weekend + Low Payoff (the trap)
Buying prompt packs · course hoarding · watching AI YouTube without building · 14 subscriptions you forgot about
The strategy in one line: do all six gold-corner moves today (about 30 minutes total), pick ONE weekend build per week, and never spend another dollar or Saturday in the bottom half.
01
Turn Memory on (2 minutes)
Claude, Settings, then look for Memory under your profile/capabilities settings and switch it on. Claude starts remembering who you are, how you work, and what you've told it, across every chat. The single best payoff-per-second toggle in AI. Then tell it three things you're tired of repeating.
02
Build your about-me.md (10 minutes)
One file that tells Claude who you are, how you work, and your hard rules. Mini version below; the full deep-dive is in my what is an .md file guide.
03
New chat per topic (0 minutes, just a habit)
Long mixed chats degrade: old context muddies new answers and burns your usage. New topic = new chat, always. With Memory and your about-me file set up, you lose nothing by starting fresh.
04
Ask Claude to push back (5 minutes)
By default, AI agrees with you more than any honest friend would. The instruction below fixes it permanently. This is the one move on this page people thank me for most.
05
Voice mode brain dumps (try it once)
Tap the voice icon in the Claude app and TALK your messy thoughts: the plan, the worry, the draft. You speak 3x faster than you type, and you're more honest out loud. Claude organizes the mess. Pairs absurdly well with a walk.
06
Pick the right model per task (1 minute to learn)
Tap the model name at the top of any chat. Daily driver for most things, the heavyweight for anything where being right matters more than being fast. My full breakdown: which Claude model to use.
The pushback instruction. Paste into Settings, then Profile, in the preferences box (it then applies to every chat), or at the top of any important conversation.
Copy the pushback instruction
Add this to your custom instructions (Claude: Settings, then Profile, "personal preferences" box) or paste it at the top of any important chat: Be my thinking partner, not my cheerleader. When I share a plan, idea, or draft: - Tell me the strongest version of the case AGAINST it before you help me execute it. - If my framing is wrong, say so before answering the question as asked. - Give me a recommendation, not a menu. Commit to a position and tell me what would change your mind. - If I'm about to repeat a mistake I've made before in our conversations, call it out. - When you're uncertain, say "I'm not sure" instead of sounding confident. Flag any claim I should verify. - Never start with praise. Start with the substance. I would rather hear a hard truth from you than a soft one from reality.
The about-me.md mini template. Fill it in once, then put it where it works: custom instructions for everywhere, or a Project's instructions for that area of your life.
Copy the about-me.md template
Save this as a note called about-me.md, fill it in, and paste it into Claude's custom instructions (Settings, then Profile) or into any Project's instructions. Ten minutes, every chat gets better. ## Who I am [Name, role, what you're building or working toward. Two sentences.] ## My life right now [The 3 things that take most of your time and attention. Update monthly.] ## How I want you to work with me - Tone: [e.g. "direct and warm, no fluff, short answers unless I ask for depth"] - Format: [e.g. "plain prose, bullets only for true lists, never tables unless asked"] - Decisions: [e.g. "recommend one option and defend it"] - Pushback: [e.g. "challenge weak ideas, I can take it"] ## Facts to always remember [The stable stuff: your family situation, your business numbers, your dietary rules, your tools, your goals with dates.] ## Never do this [Your hard rules: "never use em dashes," "never suggest waking up earlier as a productivity tip," "never invent statistics."]
The six moves above make Claude better at answering you. This part makes Claude start doing the work FOR you, and it's where the people who pull ahead actually pull ahead. The difference between someone who "uses AI" and someone who builds with it is exactly these four weekends.
Here's the ladder you're climbing, in plain words. A Project is a workspace that remembers one area of your life. A Skill is a saved playbook that fires the same way every time. An agent is a Skill with a job, a memory, and access to your real tools, basically a digital employee. And Claude Code is the workshop where the serious builds happen. One per weekend. Don't skip ahead, each week makes the next one click.
The Only Rule That Matters Here
The test for any AI Saturday: "will I have BUILT something by Sunday night?" Watching tutorials, buying prompt packs, and collecting courses all FEEL like progress and produce nothing. Building one small real thing beats consuming ten hours of content. Every single week below ends with a thing you made.
What you're really learning: the moment Claude stops being a clever stranger and starts being something that knows your actual life. Everything you build in weeks 2 through 4 stands on this foundation, so we lay it first.
Step 1, connect your tools (15 minutes). Go to claude.ai, Settings, then Connectors, and turn on Gmail, Google Calendar, and Google Drive. One click each. Now test it with a question only YOUR data can answer: "Find every recurring subscription charge in my Gmail from the last 90 days and list them with prices." When that answer comes back from your real inbox, you've felt the whole point of this week.
Step 2, build your first Project (the rest of the morning). A Project is a dedicated workspace with its own files, its own instructions, and its own memory. Open Projects, hit New Project, and build the one that pays off fastest for you. Then go deep: my 5 Claude Projects guide has the complete, fill-in-the-blank instructions for a Career HQ, a personal trainer and dietitian, a personal assistant, a money manager, and a travel planner. Build one this weekend, add the others as you need them.
By Sunday Night You'll Have
Claude connected to your real email and calendar, plus one Project that already knows who you are and stops making you re-explain yourself. This is the highest-payoff weekend on the whole list. Do not skip it to chase the shinier stuff below.
What you're really learning: how to teach Claude a task ONCE so it runs the exact same way forever. A Skill is a saved set of instructions Claude loads automatically when the task comes up. The rule the pros use: if you've done it more than once this week, it should be a Skill.
The concept, in plain words: a Skill has three parts. A name, a one-line description of when to use it, and the instructions themselves. That's it. Once it exists, you trigger it by just asking ("write my weekly update") and it runs your playbook instead of starting from a blank slate. In Claude's apps you build one under Settings, then Capabilities, then Skills (Skill Creator will even interview you and write it for you). In Claude Code, a Skill is a small file at .claude/skills/<name>/SKILL.md.
The build: pick the single most repetitive thing you do (your Monday update, your meeting-follow-up emails, formatting your content, triaging your inbox). Here's a real, fill-in-the-blank Skill you can adapt in 10 minutes. It's written for a weekly update, but the bones work for anything you repeat.
Copy this, replace every bracket with your real situation, then either paste it into Skill Creator or save it as a SKILL.md file. Then just ask Claude for your weekly update and watch it run.
Copy this Skill starter
# Save as: .claude/skills/weekly-update/SKILL.md (in Claude Code) # Or paste the instructions below into Skill Creator (Settings, Capabilities, Skills). # This example builds a weekly update. The bones work for ANY task you repeat. --- name: weekly-update description: Builds my weekly update in my exact format and voice. Use when I ask for my weekly update, status report, or Monday summary. --- You build my weekly update. Follow this exactly, every time. ## Where to get the material [Tell it where to look. Options: - "Ask me to paste my wins, tasks, and calendar for the week," or - If your tools are connected: "Read my sent email and calendar from the last 7 days and pull what I worked on."] ## Who reads it [e.g. "My manager. She skims on her phone, wants numbers first, hates filler."] ## The format (never deviate) 1. HEADLINE: one sentence, the most important true thing about my week. Lead with a number if one exists. 2. SHIPPED: 3 to 5 bullets, each starts with a verb, each includes a number where there is one. 3. IN PROGRESS: each item with a realistic finish date. Never write "ongoing" without a date. 4. BLOCKED / NEEDS A DECISION: what's stuck, why, and the one specific thing that would unstick it. 5. NEXT WEEK: top 3 priorities only. ## My voice [Paste 1 or 2 lines of how you actually write, then:] Match this tone. Plain words, confident, no corporate filler. ## Quality bar Before showing me, reread it as a skeptical reader. If anything is vague, fix it first. A wrong update is worse than a late one, so if my material is thin somewhere, ask me one question instead of guessing.
The Mindset Shift
Once you've built one, you start seeing Skills everywhere: the email you rewrite ten ways, the report you reformat every Friday, the caption style you keep re-explaining. Each one is 10 minutes to capture and saves you that task forever. Three Skills in, you'll wonder how you worked without them. Deeper walkthrough: weekend automation and what is an .md file.
What you're really learning: the difference between Claude that talks and Claude that DOES. Regular Claude tells you what to do. Claude Code reads your files, writes new ones, runs commands, and searches the web, itself, on your actual computer. This is the level most people never reach, which is exactly why reaching it puts you ahead of nearly everyone.
You do not need to be a coder. This is the part people get wrong. Claude Code is a workshop for anyone who works with files and information: organizing a messy folder, turning a pile of notes into a clean document, building a simple tracker, renaming a hundred photos by date. You describe what you want, it does the work and shows you what changed.
The build:
01
Install it. Go to code.claude.com/docs. There's a desktop-app version and a terminal version, and the docs walk you through either in about 5 minutes. It's included in your Claude subscription.
02
Open it in a folder that matters (your documents, your business files, your content). That folder becomes its workspace, and it can only touch what's in there.
03
Give it a real job, as a command not a question: "Look at this folder, build a logical structure, move everything into it, and show me what went where." Watch it actually do it. That moment reframes what AI is for.
04
Give it a memory. Create a CLAUDE.md file in that folder (a plain text file that tells it your rules and preferences once, and it reads it automatically every session). The full starter is in my 7 levels of Claude Code guide, which is the deep companion to this whole week.
By Sunday Night You'll Have
Claude Code installed, one real task it completed for you on your own machine, and a CLAUDE.md that makes every future session smarter. You'll also understand why this is the highest ceiling on the entire map: everything above level 3 lives here.
What you're really learning: how to turn everything from weeks 1 through 3 into something that runs on its own. An agent is the payoff of the whole ladder. It's not a chat you start, it's a worker you assign.
The formula, so it stops being mysterious: an agent is just Role + Context + Connections + Tools + Rules. A role (the job it owns). Context (files that teach it your world). Connections (access to your real Gmail, Calendar, data). Tools (what it's allowed to actually do). And rules (what it must never do, and when to ask you first). A Project gives it context and memory. Connectors give it the connections. A Skill gives it a repeatable playbook. Stack those together and you've built an employee.
The build: start with one job, the morning brief, because it touches everything you set up and pays off every single day. Here's a complete starter you can stand up this weekend.
Make sure Gmail and Calendar are connected from Week 1. Create a Project called "Morning Brief," paste this as its instructions, fill in the brackets, and then each morning just say "brief me."
Copy this agent starter
You are my Morning Brief agent. Every morning you pull from my connected tools and hand me one short, ranked brief so I start the day knowing exactly what matters. You are not a chatbot I question, you are a worker I assign. ROLE: My personal chief of staff for the first 10 minutes of every day. CONTEXT (fill this in once, update as life changes): - Who I am and what I do: [your role, businesses, projects] - My current top priorities, ranked: [the 3 things that actually matter right now, with any deadlines] - My VIPs: [the people whose messages always matter: boss, key clients, family] - What I want to ignore: [newsletters, promotions, automated notifications, the noise] CONNECTIONS (turn these on at claude.ai, Settings, Connectors): - Gmail, Google Calendar, and Google Drive at minimum. Add Slack or others if you use them. WHAT YOU DO each morning when I say "brief me": 1. TODAY AT A GLANCE: every meeting on my calendar with the time, and for the 1 or 2 that matter most, a one-line "here's what you need walking in." 2. WHO IS WAITING ON ME: scan my inbox for messages from my VIPs and anything that needs a reply today. Skip the noise I listed above. For each, one line: who, what they want, and how urgent. 3. WHAT'S DUE: deadlines, renewals, or commitments landing in the next 3 days that I should act on now. 4. THE ONE THING: the single highest-leverage action that, if I do it today, makes the day a win. Tie it to my ranked priorities. Just one. 5. ANYTHING UNUSUAL: a conflict on my calendar, a bill that looks off, a thread that's been ignored too long. RULES: - Short and ranked. I should read this in under 2 minutes. - Draft, never send. If you suggest a reply, show it to me, I hit send. - Never take an action that's hard to undo without asking me first. - If nothing matters in a category, say "nothing today" and move on. Don't pad it. - Learn my patterns: if I keep ignoring something you flag, stop flagging it. If I always act on something, surface it higher.
Where This Leads
Once you've built one agent, you can build a team of them: a content agent, a money agent, a research agent, each owning one job and reporting to you. That's exactly how I run two e-commerce brands and this page. The full org chart, every tool, and the co-founder agent at the center are in my AI team setup guide. You just built the first hire.
The Only AI Masterclass You Need
If this guide helped, but you’re looking to go deeper, I got you!! My 30-Day Challenge takes you from saving AI tips you never use to actually building with AI, step-by-step.
I show you exactly how I automated two e-commerce brands, my social media, and most of my personal life, then hand you the agents, workflows & systems to do the same. I’m teaching you every single thing I know with one lesson and one build a day.
Join the AI Masterclass →© 2026 Mariah Brunner. All rights reserved.