A weekly digest for Beginner Builders

⚡ TLDR

This week we show you how to set up Claude so it actually knows who you are and how you work – no more generic answers or long prompts every time. We've also built an automated morning news briefing that hits your inbox at 7am without you lifting a finger.

In the Bigger Picture: Google just gave NotebookLM a serious upgrade, and new Anthropic research reveals which jobs AI is actually coming for (the answer might surprise you). Plus two YouTube picks worth your time this week.

→ AI Skills for All: This is an oldie but a goldie from Sabrina Ramonov. Her approach is practical, no nonsense tips for all. And her lessons are timeless: be skeptical, love learning and build in public, among the many gems of wisdom. Watch here.

→ The 10 hour Claude Code Masterclass: Nate Herk has done it again. He brought us the N8N masterclass last year, and now he’s created a beast of a course. The ultimate Claude Code course that will take you from absolute beginners to highly proficient. This make take you weeks to complete, but it will be worth it. Watch here.

Join the AI-Enabled Marketer Skool community, which takes you from absolute beginner to AI-Enabled Marketer with Myosin.xyz, a community of top marketing professionals. - https://www.skool.com/the-ai-enabled-marketer/classroom

Most people use Claude like a search engine. Ask a question, get a generic answer, try again with a longer prompt. That's not how you get the best out of it.

The real unlock is Cowork, a mode inside the Claude Desktop app that gives Claude direct access to your files. Instead of chatting back and forth, you describe an outcome, Claude makes a plan, and delivers finished work straight to your folders. Documents, spreadsheets, presentations, organised files. It runs on the same technology as Claude Code, built for people who don't write code.

It only works well if you set it up properly. Here's how to do that in 30 minutes, once, and never write a long prompt again.

Before you start:

Download the Claude Desktop app at claude.com/download. Cowork is not available on web or mobile. You need a paid plan (Pro at $20/month, or Max, Team, or Enterprise). Once you're in, select Opus 4.6 as your model and turn on Extended Thinking. Then click the Cowork tab at the top. It's a different mode from Chat entirely.

Build your folder (do this once):

Create a master folder on your computer called "Claude-Work" with four subfolders inside:

Files to create inside ABOUT ME:

This is where most people skip ahead and then wonder why Claude gives generic answers. Create three small text files and save them as .md:

about-me.md — What you do day to day, not your CV. Your role, your tools, your typical projects. Think of it as an onboarding doc for a new colleague.

my-voice.md — Your tone, phrases you hate, two or three real writing samples. Paste actual examples of your work, not descriptions of how you write.

my-rules.md — How you want Claude to behave. For example: "Ask questions before executing. Show a plan first. Never delete files without my approval."

One strong file with real examples of your work beats 50 random uploads.

Set your global instructions (once, never again):

Go to Settings, then Cowork, then Edit Global Instructions. Paste something like: "I'm [Your Name], [Your Role]. Read my files before every task. Ask questions before executing. Show a plan first. Never delete files without my explicit approval."

These instructions load automatically every time you open Cowork.

What to stop doing:

Writing long prompts for every task. Your folder handles the context now. Using Chat when you need Cowork. Chat is for quick questions, Cowork is for real work. Expecting Claude to know you without files. It can't remember across sessions. The files are the memory.

Try this now: After setting up your folder, open Cowork, point it at Claude-Work, and type: "Read my ABOUT ME folder. Tell me what you know about me and how I like to work. Flag anything that's missing." That first response will show you how much better Claude gets when it has context.

Cowork is available on Mac and Windows. Requires Claude Desktop and a paid plan.

Every build so far needed you to do something. Type a message, submit a form. This one runs while you sleep.

This week's build is an automated AI news briefing. Every morning at 7am, it pulls the latest AI articles from TechCrunch, has AI pick the three most important stories and write a short briefing on each, then emails it straight to your inbox. You wake up, open your email, and you're caught up on AI news before your coffee's ready.

How it works: a Schedule Trigger fires at 7am, an RSS Feed node pulls the latest AI articles, a Limit node keeps just the five most recent, a Code node bundles them into one text block, AI reads all five, picks the top three, and writes a briefing explaining what happened and why it matters, then Gmail sends you the formatted email with links to the full stories.

Quick start:

Import the workflow JSON into n8n (link below). Connect your credentials: OpenAI on the Write Briefing node, Gmail on the Send Briefing Email node. Open the Send Briefing Email node and change the "To" address to your actual email. Click "Test Workflow" to run it now and check your inbox.

That's it. No Google Sheet to set up this time.

After running the test, check your inbox. You should have an HTML email with today's date, three AI stories with headlines and short summaries, and links to the original articles. If it looks good, toggle the workflow to Active in the top-right corner. From that point it runs every morning on its own.

Something not working? The most common issue is the Gmail node. Make sure emailType is set to HTML, not text, otherwise the formatting shows as raw code. If the AI output looks wrong, copy the Bundle for AI output and paste it into Claude or ChatGPT for troubleshooting.

Make it yours:

Change the schedule. The Schedule Trigger is set to 7am. Open the node and change the time. If you’re feeling technical, you can change the ‘cron expression’ like this: 0 6 * * * for 6am, or 0 9 * * 1-5 for 9am on weekdays only. 

Swap the news source. Any RSS feed works. Try VentureBeat (https://venturebeat.com/category/ai/feed/) or The Verge (https://www.theverge.com/rss/ai-artificial-intelligence/index.xml), or paste in any RSS URL from your industry.

Add more feeds. Duplicate the RSS node, add a Merge node to combine both feeds, then connect the merged output to the Limit node. Now AI picks the best three across multiple sources.

Change the briefing style. The system prompt says "conversational and clear, like a smart colleague." Want it more formal, more technical, or shorter? Edit the system prompt in the Write Briefing node.

Send to Slack instead. Swap the Gmail node for a Slack node and post the briefing to a team channel.

Swap the model. The default is gpt-4o-mini for speed and cost. For sharper editorial picks, try gpt-4o or Claude Sonnet.

To access the JSON you need for this build, just head on over to the AI Enabled Marketer Skool community through the link here - remember your first 7 days are free!

Google just turned NotebookLM into a content machine.

If you have not tried NotebookLM yet, this week is a good reason to start. It is a free Google tool that lets you upload documents, PDFs, or YouTube videos and ask questions about them.

Google just launched Cinematic Video Overviews in NotebookLM, using a combination of AI models including Gemini 3 and Veo 3 to generate fluid animations and detailed visuals from your source material. Alongside that, recent updates have added infographics, slide decks, flashcards, and quizzes — all generated from whatever you upload, and grounded only in that material.

Most features are free at notebooklm.google.com. The cinematic video feature requires a paid Google AI Ultra subscription.

There is a lot of noise about AI replacing jobs. This week Anthropic published research that sharpens the picture.

AI models are theoretically capable of handling 94% of tasks in computer and maths roles, yet in observed professional use, Claude currently covers only 33% of those tasks. The gap between what AI can do and what it is actually doing is enormous.

The workers most at risk are not who most people picture. The most AI-exposed group earns 47% more on average and is nearly four times as likely to hold a graduate degree. Lawyers, analysts, developers. Not trades or hospitality.

The people learning to work alongside these tools now will be best placed when that gap closes. Which is broadly why you are here.

That's it for this week.

Building something cool with AI? Reply to this email. I feature reader projects. 💬

P.S. If this was useful, forward it to someone upskilling in AI.