Let's build real agents, not just demos

A hands-on journey to building production-ready agents, not just prototypes.

I bet you know this guy 👆

Yep, Steve Jobs. The visionary behind Apple … come on, let’s be real, this is a newsletter for engineers. We should probably be rooting for Wozniak 🤣

Anyway, back to the point.

In this 1985 talk, Jobs shared a great idea: he hoped technology would evolve to the point where we could interact with the preserved wisdom of great minds like Aristotle.

Pretty wild to think about, right?

And so my hope is someday when the next Aristotle is alive, we can capture the underlying worldview of that Aristotle, in a computer, and someday some student will be able to not only read the words Aristotle wrote but ask Aristotle a question and get an answer.

— Steve Jobs, 1985

That’s a bold vision. Especially considering ChatGPT, LangChain, and multi-agent systems were just sci-fi back then.

Well, guess what, Steve—

Paul Iusztin

and I kinda made that happen …

Not just Aristotle. We’ve got Plato, Descartes, and even Turing. All brought to life as AI agents, living inside a simulated world.

Steve, meet PhiloAgents.

PhiloAgents is a collaboration between

The Neural Maze

and

Decoding ML

. For video lessons, head over to my YouTube channel (links below), and for the written deep-dives, don’t miss the incredible articles

Paul Iusztin

’s been publishing in his newsletter!

💡 Pro tip: Check out the code repository to follow the full course syllabus and get access to all the resources, articles and videos!!

---

Forget the demos. Let’s build real Agents

When Paul and I started this project, we had a clear goal in mind.

There are tons of agent projects out there—but let’s be honest, 99.99% of them are just demos. Something you could throw together in a single script.

We wanted to do something different. We wanted to show students how to actually build an agentic system like a real AI Engineer. And that’s how this course came to life.

PhiloAgents is all about that. You’ll learn how to build an AI simulation agent that brings historical philosophers to life inside an interactive game environment.

The course is broken into 6 in-depth modules, where you’ll not only learn how to build agentic apps—you’ll also architect and implement a production-ready agentic RAG system, using real LLMOps principles.

Let’s break it down, lesson by lesson 👇

---

Lesson 1: PhiloAgents architecture

In this lesson, expect more diagrams than code—and that’s on purpose.

We’ll walk you through the general architecture of the system so you understand how all the moving parts fit together. It might feel a bit high-level, but trust us: this lesson is crucial. If you skip it, the rest of the course might feel like jumping into the deep end without knowing how to swim.

Take your time here. This is where everything starts to click.

---

Lesson 2: Building a LangGraph NPC

Missed coding? Don’t worry—we’re diving in now, and from here on out, it’s all code.

In this lesson, we’ll implement the LangGraph workflow, which serves as the “brain” behind each of our Agent NPCs. You’ll see how this structure powers decision-making and interaction within the simulation.

We’ll also show you how to hook it up to one of our favorite LLM providers: Groq.

It’s free. It’s insanely fast. And hey—you’ll officially be done with GPUs.

You’re officially on the LPU team now 🤣

---

Lesson 3: Agent Memory Systems with MongoDB

As Paul likes to say, memory is the secret sauce of any good agent system—and I couldn’t agree more.

That’s why PhiloAgents dedicates a full lesson to this essential piece of agent design.

In this module, we’ll build both short-term and long-term memory systems using MongoDB as our go-to database:

• Short-term memory will be handled with regular MongoDB collections to store and manage LangGraph state.

• Long-term memory will be implemented using MongoDB’s vector search capabilities.

---

Lesson 4: Real-time Agents with WebSockets (FastAPI)

Now that your agent has a brain and memory, it’s time to connect it to the outside world.

In this lesson, we’ll show you how to expose your LangGraph application using FastAPI. But here’s the twist—we’re skipping the usual HTTP protocol and going with WebSocket instead.

Why? Because WebSocket is perfect for real-time communication between your game’s UI and the agentic backend.

This might just be one of the first resources out there showing how to use WebSocket for agent applications—so we’re pretty excited to share it with you.

Hope you like it! 🙂

---

Lesson 5: A hands-on introduction to LLMOps with Opik

It’s one of the most commonly overlooked parts of building with LLMs—but also one of the most important.

Lesson 5 is a deep dive into LLMOps:

We’re talking agent observability, prompt tracing, prompt versioning, RAG evaluation, and more.

You’ll get a solid, practical intro to everything you need to make your agentic systems reliable, debuggable, and production-ready—all using Opik.

---

Lesson 6: Bonus!

In this final lesson, the written and video content take slightly different paths—but both are packed with value.

In the written lesson, Paul delivers an amazing breakdown of every AI Engineer’s essential Python dev stack. It’s the kind of resource you’ll want to bookmark and revisit as your projects grow more complex. 👇

From 0 to Pro AI Engineering Roadmap

As for the video lesson, we put together a 2.5-hour deep-dive video that covers everything from the course—start to finish.

It kicks off with a fun intro where Paul and I introduce ourselves, and then walks you through all the key lessons in one self-contained package.

If you’re into long-form, structured content and like having everything in one place—this video’s for you.

---

Before we wrap up … just a quick note.

This isn’t an easy project. And that’s by design.

PhiloAgents isn’t meant to be a quick, copy-paste tutorial. It’s a real, complex application. It’s going to take time, effort, and a bit of struggle to get through it all—and that’s exactly the point.

Paul and I truly believe this is the only way to grow into a real AI Engineer. So if you’re feeling stuck or tempted to quit, take a breath, step back if you need to—but don’t give up.

Keep pushing, keep building.

Your future self (probably a badass AI Engineer by then) will thank you for it.

Alright, enough philosophy for now.

Hope you enjoy the course and get a ton of value out of it. I’m already working on more open-source courses for

The Neural Maze

, so stay tuned...

See you next Wednesday! 👋

---