Honestly, I feel u. I went through the exact same struggle last year when I wanted to move from web dev into AI. I started with Brilliant Premium Annual Subscription because the ads make it look so fun, right?? But tbh, I was pretty disappointed with the technical depth. It's basically a game. It's cool for high-level logic, but after finishing their neural network path, I still had NO idea how to actually code a model. It's basically all 'drag and drop' or filling in blanks, which doesnt stick when you're staring at a blank VS Code window.

So yeah, I ended up switching to DeepLearning.AI Machine Learning Specialization on Coursera and even though it's lecture-heavy, it's MUCH better for a programmer. You actually use Python and NumPy to build things from scratch. It's like $49 a month, so it's a bit of a sting if you're slow, but the hands-on labs are the only way to really learn imo. If you wanna actually BUILD models, skip the game-like stuff and go for Coursera. gl!!

Big if true

Just sharing my experience: I went through this last year as a dev too. I tried to build my own models using just interactive sites, but honestly, it felt kinda risky for actual production code. I ended up needing way more math and technical depth than I expected.

* Focus on the Fast.ai Practical Deep Learning for Coders course if you want hands-on coding.
* Check out the Udacity AI Programming with Python Nanodegree for a more structured, project-based flow.

Basically, the "fun" apps didn't give me the *rigorous* foundation I needed for real-world ML deployment. peace

> Since I can only afford one subscription right now, I’m trying to be strategic. Just saw this and figured I'd chime in. If you're deciding between the two for actually building stuff, I'd suggest Coursera, but you really need to be careful about which specialization you pick. Brilliant is nice for concepts, but it honestly won't get you to a point where you can deploy a model. You might want to consider the DeepLearning.AI TensorFlow Developer Professional Certificate. It's more of a grind, but it's very hands-on with coding. Make sure to:

• Check if the course uses TensorFlow 2.x, because older versions are a headache
• Set a strict schedule so the monthly sub doesn't drain your wallet
• Look into the Google Cloud Professional Machine Learning Engineer Certification path if you care about the ops side I'd stay away from Brilliant if you're already a dev; it feels a bit too much like a toy for someone who already knows how to code. Coursera is dry, yeah, but it's the only one of the two that'll actually show you how to handle data pipelines. Just don't get stuck in tutorial hell... try to build something alongside the videos or you'll forget it all in a week.

Honestly, I saw this earlier and wanted to chime in because most people ignore the reliability and safety aspect when starting out. Before I give my full take, what kind of models are you actually planning to build? Like, are we talking about a fun side project or something where data integrity and safety actually matter, like financial tools or medical stuff? If you want to build systems that are actually robust, you need to understand where things can break, and that's where the platform choice matters: * Coursera's Mathematics for Machine Learning and Data Science Specialization is way better for this. It forces you to grasp the 'why' behind the gradients, which is key for troubleshooting.
* Understanding the underlying math isn't just about theory; it's about knowing when your model is becoming unstable or giving biased results.
* Brilliant is cool for the vibes, but I mean... it's risky if ur trying to ship production-grade code because you might treat the AI like a black box. Basically, for reliability-first engineering, you kinda need that dry, lecture-heavy stuff to really grasp the edge cases. It's less about 'making it work' and more about 'making it safe.'

I spent a ton of time doing a "market analysis" of my own education before I committed to a subscription. Honestly, it’s all about where these brands sit in the learning ecosystem. Brilliant is basically the "Duolingo of STEM"—it’s incredible for the initial hook and building mental models, but it’s fundamentally an edutainment play. Coursera, on the other hand, is basically an extension of the legacy university system. I found that for a mid-level dev, there’s a huge middle ground that both of these kinda miss. I tried a few different paths and Educative.io Machine Learning for Software Engineers felt SO much more natural for my brain. It’s text-based with embedded environments, so you’re actually coding in a sandbox instead of watching videos or dragging sliders. It matches the dev workflow way better than a dry lecture or a game-like app. It’s about finding that sweet spot between "too easy" and "too academic." Quick tips:
1. Focus on platforms that offer live "in-browser" coding environments—if you aren't breaking actual code, the syntax wont stick.
2. Definitely look into the Hugging Face NLP Course for modern, practical LLM stuff that the legacy platforms are often too slow to update. Basically, pick the tool that matches how you actually work at your day job!

Just wanted to say thanks for everyone chiming in. Super helpful discussion.