The YouTube channels that actually teach AI product managers something. Skip the hype talkers — these creators do the work.
Why YouTube Beats Reading for Some Topics
Some concepts — attention mechanics, gradient flow, agent loops — are 10x clearer with animation than with prose. The channels below don't compete with blogs. They complement them.
We picked channels that update consistently, explain rather than narrate, and respect your time with tight editing. Watch 1.5x. Take notes. Pause to think.
🎬Watching is preparation. The AI PM Masterclass is where you ship. 4 weekends, real projects, taught by a Salesforce Sr. Director PM.
For Technical Foundations
1. Andrej Karpathy
Former Director of AI at Tesla, founding member of OpenAI. His "Neural Networks: Zero to Hero" series and his recent "Intro to Large Language Models" and "Let's build GPT from scratch" videos are the single best free AI education ever published.
Why AI PMs need this: One Karpathy video equals a CS course. Start with "Intro to LLMs" (1 hour). You'll understand the stack better than 95% of PMs. Pair with our how LLMs work guide.
Visit Andrej Karpathy2. 3Blue1Brown (Grant Sanderson)
The animated math channel. The "Neural Networks" series — including the 2024 transformer videos — visualizes attention, embeddings, and gradient descent in a way that no textbook can. Beautiful, rigorous, free.
Why AI PMs need this: When you finally see attention as a weighted sum across positions, it stops being a buzzword. Watch "But what is a neural network?" first. See our transformer architecture guide for the PM lens.
Visit 3Blue1Brown3. AI Explained
Anonymous British host who reads every paper, watches every keynote, and stitches them into 15-minute primers that are denser than most full-day conferences. The benchmark coverage is the best on YouTube.
Why AI PMs need this: You'll know about a new model's actual capabilities before your engineering team finishes scrolling Twitter. The "GPT-5", "Claude 4", and "o-series" deep-dives are required watching.
Visit AI Explained4. Yannic Kilcher
Yannic walks through ML papers section by section, on a whiteboard, with no edits and no music. The format is unglamorous and excellent. He'll cover papers the day they drop.
Why AI PMs need this: If you want to read papers but can't yet, watch Yannic do it for a few weeks. You'll absorb how to skim, where the contributions actually live, and how to spot weak claims.
Visit Yannic Kilcher5. Two Minute Papers (Károly Zsolnai-Fehér)
Short, energetic walkthroughs of impressive new research, with a focus on graphics, video, and physical simulation models. Heavy on visuals, light on math.
Why AI PMs need this: When you need to pitch what's becoming possible to a non-technical exec, this is your demo reel. "What a time to be alive!"
Visit Two Minute PapersFor Long-Form Conversations
6. Lex Fridman Podcast
Long, patient interviews with frontier figures: Sam Altman, Demis Hassabis, Yann LeCun, Elon Musk, Dario Amodei. Lex doesn't rush, which lets guests actually think on camera. Skim the transcript first; watch the parts you care about.
Why AI PMs need this: Strategy is a function of how leaders think. Watching them think out loud for 3 hours is rare and valuable. The Demis Hassabis and Andrej Karpathy episodes are essential.
Visit Lex Fridman Podcast7. Lenny's Podcast
The PM podcast. Lenny interviews PM leaders at the top AI companies — OpenAI, Anthropic, Notion, Linear, Perplexity, Cursor — about shipping AI products. Notes are detailed; clips travel well.
Why AI PMs need this: The "How AI is changing PM" episodes have specific tactics: how teams are restructuring, what tools they're using, what metrics they care about. Pair with our AI PM tool stack guide.
Visit Lenny's Podcast8. No Priors (Sarah Guo & Elad Gil)
Investor-led AI podcast with sharp, pattern-matching guests. Sarah and Elad have been building and funding AI companies long enough to ask non-obvious questions. Less performative than most VC content.
Why AI PMs need this: Investor framing is different from operator framing. You'll see how funders are bucketing the space, which moves up the priority list of every CEO you'll work with.
Visit No Priors9. Latent Space Podcast (swyx & Alessio)
The AI engineering podcast. Interviews with founders and engineers at LangChain, Modal, Cursor, Anthropic, Together, and most of the AI tooling layer. Topical, fast-paced, written show notes.
Why AI PMs need this: If you work with engineers, this is the room they're in. Watching builds your taste for what's hard, what's overhyped, and which tools to bet on.
Visit Latent SpaceFor Builder Energy
10. Matt Wolfe
Best weekly AI tools roundup on YouTube. Matt covers what's new, what works, and what to ignore. Less depth than AI Explained, more breadth across the consumer/SMB tooling layer.
Why AI PMs need this: Your users hear about AI from creators like Matt. Watching keeps you calibrated to the consumer narrative shaping their expectations.
Visit Matt Wolfe11. David Ondrej
Practical AI agent and automation tutorials. Builds and debugs real projects on camera using n8n, Claude, and OpenAI APIs. The format is hands-on rather than commentary.
Why AI PMs need this: If you've never built an agent, watching one get debugged is the fastest way to understand the failure modes. Pair with our build your first AI agent guide.
Visit David OndrejFor Strategy and Macro
12. a16z (Andreessen Horowitz)
The AI fund's video output is a mix of portfolio company interviews and frameworks talks. Variable quality — but the Marc Andreessen and Martin Casado deep-dives are sharp, and the "AI for X vertical" series is useful for strategy framing.
Why AI PMs need this: You'll absorb the dominant Silicon Valley narrative for AI markets. Even if you disagree, knowing the script matters.
Visit a16z13. The Acquired Podcast
Long, deeply researched company histories — including Nvidia, OpenAI, TSMC, and Microsoft — that explain how the AI infrastructure layer actually came to be. 4-hour episodes you'll listen to twice.
Why AI PMs need this: Strategic intuition comes from pattern recognition across industry histories. Acquired is the best free history class for tech operators.
Visit Acquired14. Stanford Online — CS25 Transformers United
Stanford's free graduate seminar, with guest lectures from Geoffrey Hinton, Andrej Karpathy, OpenAI researchers, Google DeepMind, and Anthropic. Public lectures, no paywall, full series available.
Why AI PMs need this: If you want to skip the hype layer entirely, watch academics talking to academics. Slower, harder, and far higher signal.
Visit Stanford CS2515. AI Engineer (Conference talks)
The official channel for the AI Engineer Summit and World's Fair. Keynote and workshop recordings from people building production AI at scale — eval design, RAG, agents, fine-tuning. Watch keynotes, skim the rest.
Why AI PMs need this: Production talks beat product launches. You'll see what teams are actually struggling with and how they're solving it. Pair with our best AI conferences guide.
Visit AI EngineerWatching Strategy
Treat YouTube like a textbook, not a feed. Pick three channels, watch one video per channel per week, take notes by hand. Watch at 1.5x. Skip everything that doesn't tie back to a decision you're making this quarter. Three channels, deeply absorbed, beats fifteen channels half-watched.
How to Get the Most Out of YouTube
Use YouTube Premium. Background play and downloads make this 10x more useful. Worth the cost if you commute or travel.
Watch on a real screen, not while doing dishes. Karpathy at half-attention is mostly noise. Karpathy with a notebook is a Stanford course.
Use transcripts. Most of these creators publish full transcripts. Reading is faster for parts you don't need to see animated.
Build a syllabus. Don't subscribe and hope. Pick the 5 most important videos for your current goal — say, "ship our first agent" — and watch them in order.
What to Skip
"Top 10 AI Tools" listicles. Most are affiliate-driven. The tool that's actually best for your job changes monthly; lists go stale instantly.
Hot-take reaction videos. Reading the actual paper or release notes for ten minutes beats watching 30 minutes of someone else react.
Influencer "AI agent" demos with cherry-picked outputs. If they don't show failures, assume there are many. The honest builders show both.
Build Your Watchlist in 5 Minutes
Subscribe to four: Karpathy, 3Blue1Brown, Lenny's Podcast, AI Explained. Spend 90 minutes a week on these four. You'll be technically literate, strategically current, and operationally sharp — all from free content.
When you're ready to convert input into shipped product, our AI Product Management Masterclass is the most direct path. 4 weekends, real builds, taught by a former Apple Group PM.