Hey, it’s Andreas.

I’ve said it before, but I’ll say it again: Claude Code is the most important tool I’ve seen released in years.

Lately I’m spending most of my work in the CLI with Claude Code - and I can’t remember the last time a new technology genuinely got me this excited. It feels like a real step-change in how work gets done. No matter your role or industry, this is one of those moments where it’s worth pushing yourself to try it.

In this edition, I’ll share:

• A quick setup guide to install Claude Code in 5-10 minutes

• One underrated trick: structured interviewing that produces a real plan - and helps you ship a surprisingly solid app, fast (most people don’t know this).

Plus, some of the key highlights from last week:

• Anthropic: publishing Claude’s Constitution

• Google: integrating Hume’s emotional voice tech into Gemini

• IBM: what the enterprise of 2030 could look like, based on interviews with 2,000 executives.

Let’s dive in.

Weekly Field Notes

🧰 Industry Updates
New drops: Tools, frameworks & infra for AI agents

🌀 Anthropic on publishing Claude’s Constitution → A public “values spec” for Claude: safety first, then ethics, policy compliance, then helpfulness - plus an explicit stance on Claude’s wellbeing and refusing shady requests. 

🌀 Anthropic on Claude for Excel → Claude’s Excel integration adding multi-sheet context, longer sessions, and safer edits that avoid overwriting cells.

🌀 Google on integrating Hume’s emotional voice tech into Gemini → DeepMind hired Hume’s CEO and a small engineering team under a licensing deal, aiming to improve Gemini’s voice with emotion and tone awareness.

🌀 Amazon One Medical on an agentic Health AI assistant → A new in-app assistant that uses your medical record to explain lab results, manage meds, and book care when needed. 

🌀 OpenAI on age prediction for under-18 safety → Auto-detects likely age to apply stricter defaults and reduce teen risk.

🌀 Acrobat added new main AI features → helps you to turn documents/PDFs into a presentation outline and slide deck, and can generate podcast-style audio summaries.

🎓 Learning & Upskilling
Sharpen your edge - top free courses this week

📘 DeepLearning.AI on “Gemini CLI: Code and Create with an Open-Source Agent” → A short course on building multi-step agent workflows from the command line with Gemini CLI.

📘 Anthropic on prompting best practices for Claude 4.5 → Claude now favors precise, explicit instructions over vague prompts.

🌱 Mind Fuel
Strategic reads, enterprise POVs and research

🔹 IBM on the 2030 “smarter enterprise” → After surveying 2,000+ executives, IBM’s signal is clear: by 2030 AI becomes the business model - driven by non-negotiable bets, reinvested productivity, company-specific AI moats.

🔹 Tony Blair Institute on AI sovereignty beyond the hype → The report argues sovereignty isn’t owning the full AI stack, but making deliberate choices on where to build strength and where to partner.

🔹 Runway on AI video detection hitting a tipping point → In Runway’s study, 90%+ of participants couldn’t reliably spot Gen-4.5 image-to-video clips, with only 99 of 1,000+ scoring 75%+ accuracy.

🔹 WEF Davos on AI: geopolitics, jobs, and the “no coasting” era → Amodei warned on chip exports to China and said end-to-end software work is 6-12 months out; Nadella said incumbents will “get schooled” if they can’t change fast. 

🔹 Gallup on workplace AI adoption plateauing → Nearly half of U.S. workers never use AI; adoption is 66% in remote-capable roles vs 32% in on-site roles, and leaders use it far more than individual contributors (69% vs 40%).

♾️ Thought Loop
What I've been thinking, building, circling this week

I am convinced that Claude Code is the most powerful skill you can learn in 2026 - whether you are technical or non-technical.

Claude Code is not just an AI coding tool. It is a real agent running on your machine, with direct access to your files, folders, and workflows. Anthropic could have also just called it Claude Agent.

That means it can:

• Work on large files and entire folders

• Run long, multi-step tasks that browser-based chatbots struggle with

• Act directly on your system: reading, writing, organizing, and creating

And it does not just tell you what to build. It builds with you, step by step, while you watch. Yes, it uses a CLI (Command Line Interface) - I’ll explain what that means in a minute. But no, it is not “too technical.” The complexity people imagine is not there. You get used to it much faster than you think.

Never used a CLI? Don’t worry - I’ll walk you through a 5-minute setup, then show an overlooked workflow that delivers a high-quality first outcome in two prompts.

Claude Code Installation - Get Running in 5 Minutes

Let’s get Claude Code installed. You'll need the following prerequisites:

• A Claude subscription (Pro at $20/month minimum)

• Node.js (install here if you don’t have it)

CLI vs Desktop vs Cloud - pick the right interface

Claude is available in three forms:

• CLI (Terminal) - fastest and most powerful

• Desktop app - convenient, but more memory-heavy

• Cloud (claude.ai) - great for brainstorming and mobile use

I use all three:

• CLI for deep work, coding, debugging, building features

• Desktop when I want Claude in a separate window

• Cloud/mobile for planning and ideation on the go

The beauty is they all share your Claude subscription. Same account, same usage limits, different interfaces for different contexts.

For this guide, we focus on the CLI (terminal) - this is where the real leverage is. The entire setup takes about 5 minutes, start to finish. I keep this intentionally lightweight - no plugins, no skills, no claude.md.

What is the CLI?

The CLI (Command Line Interface) lets you control your computer by typing commands instead of clicking buttons. It runs inside a terminal - usually a plain black window with text, no menus, and no visible safety net. That’s why it looks intimidating at first.

Developers traditionally use the CLI to:

• Install and update tools

• Run projects and scripts

• Start servers and debug code

For years, the terminal was gated by syntax. If you didn’t know the commands, you were out. Now you don’t need to speak “command language” to access the power.

With Claude Code, you simply open the terminal and use natural language. You describe what you want, and the agent translates that intent into the right commands. That single change removes most of the friction that made the CLI feel inaccessible.

Step 1: Open the Terminal

macOS
Cmd + Space → type “Terminal” → Enter

Windows
Windows key → type “Windows Terminal” or “PowerShell” → Enter

Step 2: Install Claude Code

Open your terminal and enter the following command:

macOS, Linux, WSL:

curl -fsSL https://claude.ai/install.sh | bash

Windows PowerShell:

irm https://claude.ai/install.ps1 | iex

Windows CMD:

curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

Step 3: First Launch (Claude Code)

Now, just run the following command:

claude

After the first run, you will:

• Choose a theme

• Connect your Anthropic account

• Trust the current directory

Once done, you are inside Claude Code. Which looks like below. If you run into problems have a look here.

Step 4: Start Building

That’s it. You’re up and running.

You can now:

• Chat with Claude

• Ask it to research or analyze files

• Generate code, documents, summaries

• Build apps directly into the folder

Use for example the following prompt to build your first app:

“Create a simple to-do list app using HTML, CSS, and JavaScript. Keep it in a single file.”

The Hidden Superpower: AskUserQuestionTool

Almost nobody talks about this.

Claude Code has a built-in tool called AskUserQuestionTool. When invoked, Claude does not blindly execute your vague idea. It interviews you. This works best when you start with a very short, high-level description of what you want to build. From there, Claude will ask you questions, fill in the gaps, and only then start acting.

This is the prompt to trigger the tool:

“Interview me in detail using the AskUserQuestionTool about technical implementation, UI/UX, concerns, and tradeoffs.”

Without this tool, Claude might ask one or two clarification questions about your rudimentary idea and then start building. With it, you receive deeper, structured clarification across multiple rounds before anything happens.

Especially for beginners, this is one of the best ways to understand how Claude Code actually works and to get high-quality output from the start. Planning is not overhead here - it’s how you create good context. And better context consistently leads to better results.

What’s next?

Start building and experimenting. Once you go deeper, there’s a lot more power waiting - plugins, Claude Skills, MCP, feedback loops, and more. It’s a real rabbit hole. If you’re not careful, you’ll look up and it’s midnight.

I’ve seen people use Claude Code to hack a smart oven for remote preheating, build custom Apple Health dashboards, and even automate parts of growing tomatoes. The range is wild.

And remember: Claude Code is not just for code. You can use it to organize messy folders, analyze complex data (even spreadsheets), and bring structure to work that usually feels scattered.

Enjoy!