#2 Edition: The AI Agent era is here — Here's how to get ahead

Hey, it’s Andreas.

Welcome back — and a big shoutout to the 3,315 new subscribers who joined us this week. 🙌 

Let’s dive into this week’s edition. Here’s what’s inside:

• Mistral, Perplexity, and Microsoft all make major agent moves — from orchestration APIs to full workbenches and human-agent UX

• My deep dive: A practical AI Agent learning roadmap — from zero to deployment

• and much more…

Let’s dive into it!

P.S. I changed the structure a bit today — shorter, more focused. Hit reply and let me know what you think.

Weekly Field Notes

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

🌀 Mistral launched its Agents API — their big move into the AI agent race.
→ Enables real-time search, tool use, and multi-agent orchestration.

🌀 Perplexity launched Perplexity Labs — think of it like an agentic workbench: you give it a project, and it plans, codes, browses, and builds charts.
→ Still early, but I really enjoyed playing around with it. You can try it if you have Perplexity Pro — feels like a mix between Manus, Cursor, and Deep Research.

🌀 Opera just launched Neon — a new browser that automates web tasks, creates content with AI agents, and lets users code with natural language.
→ Claims to be the “world’s first AI agentic web browser.” There’s a waitlist for now.

🌀 Microsoft launched Magnetic-UI —a human-centered web agent. If you’re curious about human–agent collaboration, this one’s worth exploring.
→ Feels experimental, but there’s a solid tutorial. Useful if you’re building agentic UIs or testing multi-modal control flows.

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

🟦 Hugging Face - MCP Course
A deep dive into the Model Context Protocol — theory, design, and implementation.
→ Technical but beginner-friendly. Great if you're serious about building agent frameworks.

🟦 IBM x DeepLearning.AI - Agent Communication Protocol (Coming soon)
How AI agents communicate and collaborate. The course is already filmed and recorded, but not released yet.
→ Launches June 25. Mark your calendar.

🟦 Anthropic - AI Fluency Course
12 short modules (3–4h total) on working with AI systems.
→ Covers delegation, prompting, discernment. Perfect for teams or structured self-study.

🌱 Mind Fuel
Strategic reads, enterprise POVs and research

🔹 IBM report on how Agentic AI is hitting core ops — Finance, HR, Procurement, Order-to-Cash, Customer Service, and Sales Support.
→ Good read about the impact and new operational models currently evolving within big enterprises.

🔹 OpenAI just published new guidance on building real-world multi-agent systems → Practical, technical — a must-read if you're orchestrating agents (fyi: all the best practices are on page 11).

🔹Cloud Security Alliance released the first Red Teaming Guide for Agentic AI.
→ Focused on real risks like permission escalation. Rare to see this — and we need much more work around risks and mitigation strategies.

🔹 Stack AI released a paper around 25 real-world agent use cases mapped across industries.
→ Great reference doc if you're looking for implementation inspiration.

🔹 a16z and Rick Rubin released a good talk on “the way to [vibe] code”.
→ A “Taoist” take on coding and creativity with AI. Weird, beautiful, refreshing. Please take it with a grain of salt.

🔹 Y Combinator published a good talk on the state-of-the-Art of prompting for AI Agents.
→ Prompting isn’t dead — it’s evolving. YC breaks down real failures, testing strategies, and production best practices.

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

You’ve probably seen it — the viral Axios interview where Anthropic’s CEO warned that “AI could wipe out half of all entry-level white-collar jobs and push unemployment to 10–20% in the next five years.”

Here’s the hard truth: It’s already happening.

AI isn’t just assisting anymore — it’s replacing. And what we’re seeing now? It’s not the ceiling. It’s the floor.

And with the next wave of AI Agents already underway, you can probably feel how big this shift really is. While the downside is obvious, the upside is massive — but only if you move now.

You don’t need to have a technical background — though it definitely helps. But you do need to know how agents work — and even more important learn how to make them work for you.

I’ve spent the past 10 months deep in the weeds: 200+ hours of content, 50+ programs, from no-code tools to full-stack frameworks.

Here’s what I’ve learned: There’s no single course that takes you from zero to building real, scalable AI agents. But the good news? You can build your own AI roadmap — completely free — with expert-led courses.

Your Learning Path (No Matter Your Background) ⬇️

Step 1: Get your fundamentals right
Don’t jump into LLM orchestration without basics.
→ Start with an AI/GenAI crash course or a beginner-friendly ML intro to build clarity and vocabulary.

Step 2: Learn automation basics
Before agents, you need to understand (AI) workflows.
→ Use tools like n8n, Make, or Zapier to build automation muscle: triggers, flows, APIs, logic.

Step 3: Start building with real agent frameworks
Once you’ve built a foundation in AI and automation, it’s time to go deeper — into the agent stack.
→ Learn with LangGraph, CrewAI, smolagents, AutoGen, LlamaIndex, and others. These frameworks help you design, manage, and scale real agent systems — from side projects to enterprise-grade solutions. Let’s break down the nine best courses I’ve come across.

Beginner-Level AI Agent Courses (Start Here 🚀)

Builder-Level AI Agent Courses (Get Practical 🛠️)

Advanced-Level AI Agent Courses (Expert Mode 💡)

🔧 Tool Spotlight
A tool I'm testing and wachting closely this week

IBM just relaunched its agent platform — and I’ve been testing it this week. The UI/UX is clean, intuitive, and feels built for real workflows.

You can try it free for 30 days — no credit card, no strings. Definitely worth checking out if you’re serious about agent orchestration and you look for a no-code or low-code environment.

Highlights:
• Visual agent builder + catalog of prebuilt assistants (100+)
• Real multi-agent orchestration (no more micromanaging bots)
• Designed for cross-business and functions and easy to plug into your stack
→ Explore watsonx Orchestrate