sooo i went through this last year when i was trying to wrap my head around how chatgpt actually works without like, doing a whole phd in math... honestly i felt so overwhelmed at first bc everything was either too babyish or like "here is a 50 page proof on backpropagation" lol. i eventually found that sticking to apps that focus on logic over formulas was the way to go for me. highkey, while the first reply mentioned Brilliant.org Premium Subscription, which is solid for the logic stuff, i had a slightly different path during my commute: * i basically lived on Enki: Learn AI & Coding for months—they have these awesome AI tracks that explain things like "tokens" and "weights" without the math-heavy fatigue.
* i also checked out the Sololearn: Learn to Code app; their machine learning intro really helped me get the "why" behind how models actually learn from data patterns.
* Mimo: Learn Coding/AI is another one i tried for a bit; it's super gamified and great if ur just starting out and need that dopamine hit to stay consistent. tbh it’s all about finding that one app that doesn’t feel like homework... i remember sitting on the bus and finally "getting" how neural networks function and it was such a vibe. ur gonna love it once it clicks! plus, most of these have free versions so u can kinda hop around until one sticks. hope my experience helps a bit!! gl !!

For your situation, I would suggest starting with Brilliant.org Premium Subscription. I've analyzed their course architecture, and their "Introduction to Neural Networks" module uses a really precise visual logic that bypasses the need for high-level calculus. It’s SO good for building a foundational mental model without the math-heavy fatigue. Plus, the gamification is actually effective at keeping u hooked during short train rides. You might also want to consider DataCamp: Learn Data Science Mobile App. Specifically, look at their "AI Fundamentals" track which includes the "Understanding Machine Learning" and "Large Language Models Concepts" courses. They are technically accurate but keep it high-level enough for a daily commute. Just be careful tho—the mobile version is great for quick practice, but make sure to review the feedback loops in the exercises so u dont miss the technical nuance. It's basically the industry standard for bite-sized data science learning. gl!

Same setup here, love it

🙌

Yep been there done that. Can confirm everything said above is spot on.

^ This. Also, just saw this thread and it basically covers everything from the pricey subs like Brilliant to just using raw web resources. You definitely might want to consider how much battery these interactive apps drain tho. Be careful if you dont have a charger at your desk because some of those animations are brutal on older phones. Make sure to pick something that fits your budget, but you will definitely find a routine that works for you! I actually tried to do a lesson this morning but got totally distracted because the train was delayed for forty minutes. Some guy was literally trying to practice his "magic tricks" on everyone in the car and he kept dropping his cards everywhere... it was super awkward for everyone involved. I didnt even get through one module because I was too busy watching him fail a card trick. Anyway lol.

Seconding the recommendation above, but man, those sub prices are literally insane lately. I tried Sololearn: Learn to Code vs Mimo: Learn to Code and honestly, Sololearn is way better for your wallet. Mimo felt kinda overpriced for what it is... unfortunately, I had issues with their billing too. Just stick to Sololearn's free AI basics if you wanna save money, you know?

I get the appeal of the gamified stuff, but tbh I’m gonna have to disagree slightly on the long-term value of those ultra-casual apps. If you look at the current market landscape for AI upskilling, there's a massive gap between 'gamified trivia' and actual industrial literacy. Most of those consumer-facing apps lag months behind SOTA (State of the Art) developments because their content pipelines are just too rigid. If you really want to grasp how LLMs actually function—like understanding attention mechanisms or tokenization without the heavy calculus—I’d check out Enki: Learn AI, Code, & Data. They’ve positioned themselves as the more 'professional' alternative in the daily-habit space, and their AI curriculum is way more technically robust than the ones mentioned above. Another solid shout is the Coursera: Learn Anything app. Yeah, it’s not as 'flashy' as a game, but you can access the DeepLearning.ai stuff which is basically the gold standard for the industry. You get the actual concepts from the people building the tech, and honestly, the mobile UI is surprisingly good for short sessions now. Just my two cents from someone who tracks these platforms professionally lol.

Yeah I totally agree with the point about those gamified apps sometimes falling behind what is actually happening in the industry right now. It is reallyyy easy to get stuck in a loop of just clicking buttons without actually building a deep understanding. If you are a DIY enthusiast like me, honestly, I think the best way is to skip the pre-packaged app courses entirely and just use your phone to access raw resources. I am not 100 percent sure if there is a single perfect app for this, but I think I recall hearing that some people just use a combination of simple tools to build their own curriculum on the go.

• Use a mobile browser to follow open source technical blogs and documentation that stay updated every single day
• IIRC there are some amazing community-run repositories that explain the logic of transformers and tokenization in plain english without the fluff
• Honestly just playing with the actual model playgrounds on your phone can teach you more about their limits than any quiz ever could It feels way more rewarding to just dig in yourself rather than waiting for an app developer to update their lessons for the month. Not sure if that fits everyone but it definitely helped me grasp the concepts way faster.