The Best AI YouTube Channels for Education
ZarifYouTube is where I learned 80% of what I know about AI. Books are slow. Courses are expensive. The right YouTube channels give you world-class teaching for free — if you know which ones to watch. Below are the channels I actually recommend.
TL;DR
- 3Blue1Brown and StatQuest are the best for ML fundamentals.
- Andrej Karpathy's channel is the single most valuable resource for understanding LLMs from first principles.
- Two Minute Papers and AI Explained are the best for staying current with research.
- Matthew Berman, Wes Roth, and David Shapiro lead the news-and-product space.
- Nick Saraev, Cole Medin, and Liam Ottley dominate the AI agents and automation tutorials.
How I Picked These
YouTube is full of AI channels run by people pretending to be experts. The list below has three filters:
- Does the host actually understand the material, or are they reading off ChatGPT-generated scripts?
- Is the content evergreen or pure news churn? Some news channels are still useful, but I weight teachers higher.
- Production quality matters less than clarity. A whiteboard and a real explanation beats a cinematic intro any day.
This isn't a "who has the most subscribers" list. It's the channels I actually open in a tab and learn from.
Best Channels for ML Fundamentals
If you're trying to actually understand how the math and architecture work, start here.
1. 3Blue1Brown
- Host: Grant Sanderson
- Subscribers: 8.2M+
- Cadence: Monthly-ish
- Best for: Building deep visual intuition for math and ML
- Why it's worth it: Grant's neural networks playlist is the single best introduction to deep learning ever made. His more recent series on transformers and attention is on the same level — built using his Manim animation library. If you don't get neural networks "in your bones" yet, watch this first.
2. StatQuest with Josh Starmer
- Host: Josh Starmer
- Cadence: Weekly-ish
- Best for: Statistics and ML fundamentals from scratch
- Why it's worth it: Josh's "BAM!" energy is divisive but his teaching is unmatched for fundamentals. Every concept gets the careful, slow, no-prerequisites treatment. Random forests, gradient boosting, transformers — all explained without losing you.
3. Andrej Karpathy
- Host: Andrej Karpathy (OpenAI founding member, ex-Tesla AI lead)
- Subscribers: ~990K (approaching 1M)
- Cadence: Sporadic — when he posts, drop everything
- Best for: Understanding LLMs from the inside
- Why it's worth it: Andrej's "Let's build GPT from scratch" and "Let's build the GPT tokenizer" videos are the single best free education on how language models actually work. His "Intro to LLMs" talk is required viewing for anyone working with AI. There is no substitute.
4. Yannic Kilcher
- Host: Yannic Kilcher
- Subscribers: 316K+
- Cadence: Weekly-ish
- Best for: Paper walkthroughs at the research level
- Why it's worth it: Yannic reads papers and explains them — slowly, thoroughly, with good intuition. If you want to understand what's in DeepSeek-R1 or Llama 4, he'll walk you through the actual paper. Best for intermediate-to-advanced viewers.
Best Channels for AI News and Research
These channels keep you current without you having to read every paper.
5. AI Explained
- Host: Philip
- Cadence: 2-3x per week
- Best for: Sober, well-researched AI news
- Why it's worth it: Philip is the antidote to AI hype YouTube. Every video is well-sourced, cites papers and primary sources, and treats the audience as intelligent adults. If you want to know what actually happened with a model release, watch his take.
6. Two Minute Papers
- Host: Dr. Károly Zsolnai-Fehér (TU Wien)
- Subscribers: 1.77M+
- Cadence: 2-3x per week
- Best for: A quick, enthusiastic look at new AI research
- Why it's worth it: Short videos covering one paper at a time, mostly in graphics, robotics, and generative AI. Karoly's "What a time to be alive!" energy is part of the charm. Great for staying broadly aware of cool research.
7. Matt Wolfe
- Host: Matt Wolfe (founder of FutureTools.io)
- Subscribers: 694K+
- Cadence: 2-3x per week
- Best for: AI tool reviews and weekly news roundups
- Why it's worth it: Matt runs the most comprehensive AI tool coverage channel on YouTube. His weekly news rundowns synthesize everything that happened in 20-30 minutes. Pairs with his FutureTools directory and 225K+ newsletter.
8. Matthew Berman
- Host: Matthew Berman
- Cadence: Daily
- Best for: Hands-on demos of new AI tools and models
- Why it's worth it: Matthew tests every new model and AI tool the day it drops. Less theory, more "let's try it." Useful for getting a quick sense of what new releases can actually do.
9. Wes Roth
- Host: Wes Roth
- Cadence: Daily
- Best for: Big-picture AI news and frontier developments
- Why it's worth it: Wes is good at synthesizing across multiple sources and giving the "what does this mean" framing. Higher hype level than AI Explained, but still informative.
Best Channels for AI Agents and Automation
If you're building with AI — n8n, agents, workflows — this is where you'll spend most of your time.
10. Nick Saraev
- Host: Nick Saraev
- Cadence: Weekly
- Best for: AI agency, automation business, and serious agent builds
- Why it's worth it: Nick has built and sold real automation businesses. His content goes deep on workflows, client acquisition, and how to actually monetize AI automation skills. One of the most useful practitioners on YouTube.
11. Cole Medin
- Host: Cole Medin (Founder of Dynamous AI)
- Subscribers: 185K+
- Cadence: 2-3x per week
- Best for: AI agent architecture, tooling, and dev workflows
- Why it's worth it: Cole's content on agent frameworks, MCP, and agent orchestration is the best on YouTube right now. He's both technical and practical — code on screen, working examples, no fluff. Creator of Archon, the open-source meta-agent (13.6K+ GitHub stars).
12. Liam Ottley
- Host: Liam Ottley (Founder of Morningside AI)
- Subscribers: 713K+
- Cadence: Weekly
- Best for: AI agency and chatbot business builders
- Why it's worth it: Liam built an AI agency to scale (40+ employees at Morningside) and now teaches others. His content covers the business mechanics of selling AI services — sales, pricing, positioning. The AI Automation Agency Hub on Skool has 280K+ members. Practical for anyone running an AI consultancy.
13. Nate Herk
- Host: Nate Herk (Founder of Uppit AI, n8n Certified Expert)
- Subscribers: 595K+ (one of the fastest-growing AI channels on YouTube)
- Cadence: 2-3x per week
- Best for: n8n workflows and AI automations for business
- Why it's worth it: Nate has the cleanest n8n tutorials on YouTube. Grew from zero to 230K subs in nine months after his first video in September 2024. If you're trying to learn n8n with AI specifically, his channel is the fastest path.
14. Fireship
- Host: Jeff Delaney
- Subscribers: 4.1M+
- Cadence: Weekly
- Best for: Fast, hilarious, dev-focused breakdowns of new AI tools and models
- Why it's worth it: When a new model drops, Jeff has a 5-minute "100 Seconds of" or news take by the next day, with code samples and honest assessment. The funniest, fastest, highest-density AI dev channel on YouTube.
Best Channels for Specific Tools and Frameworks
When you need to learn a specific tool fast.
15. James Briggs (Pinecone)
- Host: James Briggs
- Cadence: Weekly
- Best for: Vector databases, RAG, and applied LLM engineering
- Why it's worth it: James is the best teacher of RAG and embeddings on YouTube. His tutorials are clean, full of working code, and assume you actually want to ship something.
16. Sam Witteveen
- Host: Sam Witteveen
- Cadence: Weekly
- Best for: LangChain, agents, and emerging frameworks
- Why it's worth it: Sam covers new tools quickly with working code in Colab notebooks. If a new framework or model just dropped, Sam probably has a tutorial on it within a week.
17. Riley Brown
- Host: Riley Brown (co-founder of Vibecode.dev, raised $9M)
- Subscribers: 221K+ on YouTube, 1.5M+ across all platforms
- Cadence: 2-3x per week
- Best for: Vibe coding, AI app building, and creator-side AI use
- Why it's worth it: Riley is the best at the "non-engineer builds with AI" angle — Claude Code for mobile apps, Cursor, Replit. Posted the first TikTok about ChatGPT the day it launched (20M views). Important for understanding where this all goes.
Best Channels for AI in Specific Verticals
18. David Shapiro
- Host: David Shapiro
- Cadence: 2-3x per week
- Best for: AI safety, AGI predictions, and post-labor economics
- Why it's worth it: David is opinionated and prolific. His takes on AGI timelines, post-labor economy, and AI alignment are unique and worth engaging with whether or not you agree.
19. AI Engineer
- Host: Conference content + interviews
- Cadence: Variable
- Best for: Talks from the AI Engineer Summit and World's Fair
- Why it's worth it: This channel hosts the talks from swyx's AI Engineer events. The talks from people at OpenAI, Anthropic, and frontier startups are some of the highest-signal content on YouTube.
How I Actually Watch
A few habits that keep this manageable:
- Subscriptions tab only — never the algorithmic home feed
- 2x speed for almost everyone except 3Blue1Brown and Karpathy (the visuals matter)
- Pocket or Notion clips — when something's worth re-watching, save it
- Build along — for tutorials, code as you watch. Passive watching is mostly wasted time.
YouTube can either be the best AI education on the internet or the biggest time sink. The difference is curation. Subscribe to the channels above, ignore the rest, and you'll be in the top 1% of AI-informed people in six months.
Channels I've Outgrown
Several big-subscriber AI channels have shifted from teaching to "react to today's headline" content. Useful early on, but eventually you graduate. The tell: when the videos stop teaching you anything new and just summarize tweets you've already read. Be honest about whether you're still learning.
FAQ
What's the best AI YouTube channel for total beginners?
Start with 3Blue1Brown's neural network playlist for fundamentals, then watch Andrej Karpathy's "Intro to Large Language Models" talk. After those two, you'll have a stronger AI foundation than 90% of people. Then pick AI Explained for ongoing news.
What's the best YouTube channel for learning to build with AI?
For agents and automation: Cole Medin and Nick Saraev. For RAG and embeddings: James Briggs. For n8n specifically: Nate Herk. For LangChain and frameworks: Sam Witteveen. Combine two or three of those depending on your stack.
Are AI YouTube channels better than paid courses?
For most learners, yes. The combination of 3Blue1Brown plus Karpathy plus a builder channel like Cole Medin is genuinely better than most paid courses — and free. Paid courses make sense when you need structure, accountability, or a credential, not because the information is better.
How do I avoid AI hype YouTube?
Three rules. Subscribe manually instead of relying on the algorithm. Unsubscribe from anyone whose thumbnails use shocked faces and red arrows. And weight channels by host expertise — a channel run by an actual researcher or builder beats a channel run by a content creator who picked AI as a niche.
Should I watch AI YouTube on 1x or 2x speed?
Almost everything works at 1.5x or 2x. Exceptions: anything with heavy math visuals like 3Blue1Brown, anything where you're coding along like Karpathy's tutorials, and anything in a language you're less fluent in. For most news and commentary, 2x is the right default.
The channels above cover the full spectrum — fundamentals, news, building, business. Pick three or four to start, kill the algorithm's grip on your feed, and treat YouTube as a curated learning environment instead of an entertainment platform. Done right, it's the best AI education available.
