Key takeaways
- Start with three channels: The AI Daily Brief or AI Explained for news, AI Engineer for practice, and Dwarkesh Patel or Latent Space for interviews.
- Subscriber count is a weak signal of usefulness. No Priors has 101K subscribers and some of the best executive guests; 3Blue1Brown has 8.76M and is mostly a math channel.
- The free conference talks on AI Engineer and the full lectures on Stanford Online replace a surprising share of paid training for engineering teams.
- Daily news channels build awareness. Verify any claim that changes a roadmap against the lab's own release notes or the paper.
- Andrej Karpathy uploads rarely, but his general-audience LLM lectures are the best single primer for executives who never studied machine learning.
Which AI YouTube channels are worth following?
Ranked within each group by how much a technology leader gets out of it. "Checked" says where each number came from and when.
News and analysis
-
459K subscribers · 2 to 4 uploads a month
For: CTOs and CAIOs who want one careful read on each frontier model release
The channel covers new models, lab statements and safety debates with more source reading than most news channels, and its author runs the Simple Bench reasoning benchmark. Uploads are less frequent than the daily shows, which works in its favour: each video tends to be the considered take rather than the first one.
Checked: YouTube channel page, 19 Nov 2026 (About links checked)
-
588K subscribers · near-daily uploads
For: Executives who want a 20-minute daily news habit with a business framing
Nathaniel Whittemore's show turns the day's AI news into a business story: enterprise adoption, agents at work, vendor moves. It also runs as a podcast and newsletter, so it fits a commute as easily as a desk. I'd pick it over the hype-driven news channels for anyone who has to brief a board.
Checked: YouTube channel page, 19 Nov 2026; sponsor email from aidailybrief.ai, 19 Nov 2026
-
147K subscribers · daily live show
For: Technology executives who track the business side of AI: funding, markets, founders
TBPN bills itself as technology's daily show, and its guest list leans toward founders, investors and operators reacting to the news in real time. Watch it for market sentiment and who is saying what, then check the claims elsewhere. The full episodes are long; the clips are the efficient way in.
Checked: YouTube channel page, 19 Nov 2026
Building with AI
-
1.04M subscribers · daily uploads
For: Engineering leaders and staff engineers building with LLMs and agents
The channel publishes the talks from the AI Engineer conferences for free, so you get practitioners from the labs, tool vendors and product teams explaining what works in production. It is the closest thing to a free conference pass for agentic engineering. Filter by speaker company and pick the talks that match your stack.
Checked: YouTube channel page, 19 Nov 2026; sponsorship email from ai.engineer, 19 Nov 2026
-
1.77M subscribers · 17 videos, rare uploads
For: Engineers and technical executives who want to understand how LLMs work from first principles
Only 17 videos, and the latest is about a year old, but the back catalogue is the point: building GPT from scratch, the tokenizer, reproducing GPT-2, and the long-form Deep Dive into LLMs and How I use LLMs. The general-audience lectures are the best single investment of three hours a non-specialist executive can make. The coding series is for engineers who want to build the intuition by hand.
Checked: YouTube channel page, 19 Nov 2026 (videos tab checked)
-
142K subscribers · several uploads a week
For: AI engineers and architects tracking new models, local AI and agent tooling
A Google Developer Expert for machine learning since 2017, Witteveen reviews new open and closed models, agent frameworks and local hardware with a builder's eye. The videos are short enough to keep pace with releases and specific enough to decide whether a model deserves a test in your stack.
Checked: YouTube channel page, 19 Nov 2026
-
154K subscribers · roughly weekly uploads
For: Senior engineers and engineering managers moving teams toward agentic coding
The channel is explicitly about becoming an agentic engineer: coding agents, multi-agent setups and what it calls software factories. The thumbnails are loud, but the content is hands-on workflow design. Useful for a VP of Engineering who wants to see what an aggressive agent workflow looks like before writing policy for one.
Checked: YouTube channel page, 19 Nov 2026
Research explained
-
1.84M subscribers · 2 to 3 uploads a week
For: Busy leaders who want a weekly sense of what AI research can now do
Short, enthusiastic summaries of new research papers, heavy on graphics, simulation and generative video. You will not learn the method in depth, but you will see results months before they become products. Treat it as a radar, then go to the paper for anything that matters to your roadmap.
Checked: YouTube channel page, 19 Nov 2026 (About page checked)
-
221K subscribers · several uploads a month
For: Technical leaders who want depth on reasoning, AGI debates and the science behind the models
MLST describes itself as a highly technical AI podcast covering research, science and philosophy, and it delivers long conversations with researchers that do not simplify for a general audience. It publishes transcripts and a paper archive alongside the videos. Pick it when you want to understand the arguments behind a capability claim.
Checked: YouTube channel page, 19 Nov 2026 (About links checked)
-
8.76M subscribers · irregular uploads
For: Anyone who needs to explain transformers and attention to a non-technical board
Grant Sanderson's math channel is not an AI channel, but its Neural networks playlist, including the chapters on transformers and attention and the eight-minute Large Language Models explained briefly, is the clearest visual explanation available. Send the short one to your board and the full series to your product managers. The channel does not take sponsorships.
Checked: YouTube channel page, 19 Nov 2026; contact policy from 3blue1brown.com FAQ, 19 Nov 2026
Lectures and explainers
-
1.35M subscribers · daily uploads
For: Engineers and leaders who want university-grade lectures on agents, evals and post-training
Stanford posts full course lectures and guest talks, with recent uploads covering AI agents, MCP and agent safety, agent evaluation and post-training methods. It is free graduate-level material on exactly the topics engineering teams are wrestling with. Pick a course playlist rather than browsing; the channel publishes too much to follow in full.
Checked: YouTube channel page, 19 Nov 2026 (videos tab checked)
-
1.85M subscribers · uploads Sunday to Friday
For: CIOs, architects and managers who need clear 10-minute explainers on AI concepts
IBM's experts explain AI, security and data topics at a whiteboard: RAG, distillation, agents and observability, software supply chain security. The vendor angle is light in most explainers. This is the channel I'd hand to a manager who needs to understand a concept before a vendor meeting.
Checked: YouTube channel page, 19 Nov 2026
Interviews
-
1.48M subscribers · weekly to fortnightly episodes
For: CTOs and CAIOs who want to hear lab leaders and researchers questioned in depth
Deeply researched long-form interviews with AI researchers, lab leaders, economists and historians. Patel prepares harder than most hosts, so the conversations get past talking points on scaling, timelines and AI risk. Best for strategic context rather than anything you will apply on Monday.
Checked: YouTube channel page, 19 Nov 2026 (About links checked)
-
543K subscribers · several episodes a week
For: AI engineers and engineering leaders who want builder interviews tied to what ships
The podcast and newsletter for AI engineers talks to founders, researchers and tool builders about models, agents and infrastructure, with a bias toward what you can use at work. It sits between research depth and product news. If your team only follows one interview show, make it this one.
Checked: YouTube channel page, 19 Nov 2026; partnerships email from latent.space/about, 19 Nov 2026
-
101K subscribers · weekly episodes
For: Executives who want the investor and CEO view of where AI markets are heading
Investors Elad Gil and Sarah Guo interview founders, CEOs and CTOs about AI markets, disruption and research. Recent guests include security, fintech and drug-discovery leaders, which makes it useful for seeing how AI changes industries outside software. The channel is smaller than its guest list suggests.
Checked: YouTube channel page, 19 Nov 2026
Which AI YouTube channels are worth an executive's time?
The channels worth an executive's time are the ones that shorten decisions, and subscriber count says little about that. For a CTO or CAIO, that means a short daily or weekly news habit, one source of practitioner detail, and one interview show that explains where the labs and markets are going. Everything else is optional.
My default stack for a technology executive is The AI Daily Brief on weekdays, AI Engineer talks filtered to the topics on the current roadmap, and Dwarkesh Patel for strategic context. That is roughly three to four hours a week, which is less than one conference day and more current than most paid briefings. If the daily cadence is too much, swap The AI Daily Brief for AI Explained, which publishes less often and tends to go deeper on each release.
Chief AI Officers have a slightly different need. They spend more time with boards and business units, so No Priors and TBPN are useful for the investor and founder view, and 3Blue1Brown's short LLM explainer is worth keeping as the link you send to a director who asks how these models work. The CAIO guide covers the rest of that role, and the CAIO archetypes page draws on TBPN's own guest list.
Which channels should engineering leaders give their teams?
Engineering teams get the most from channels that show working systems. AI Engineer is the clear first pick: it publishes conference talks from lab engineers, tool vendors and product teams, many of them on coding agents, evals and production reliability. Sharing two or three talks a sprint, tied to a real decision the team faces, works better than telling people to subscribe.
Andrej Karpathy's coding series is the foundation for engineers who want to understand what happens inside a model, and the time pays back when the team has to debug tokenizer issues, context limits or odd model behavior. Sam Witteveen and IndyDevDan cover the faster-moving layer: new models, agent frameworks and workflows. IndyDevDan in particular shows what an aggressive agentic workflow looks like, which is useful input before you set policy on agentic coding or AI code review.
Stanford Online is the underused option. Its recent uploads include full lectures on AI agents, MCP and agent safety, agent evaluation and post-training methods, which is graduate-level material on exactly the questions teams are arguing about. For structured learning beyond YouTube, compare the paid options on the agentic AI courses page.
How should you separate signal from hype on AI YouTube?
AI YouTube rewards speed and excitement, and both work against accuracy. The fastest channels publish within hours of a release, often before anyone has tested the model on real work. That is fine for awareness, but it is a poor basis for a vendor decision or a reorganisation.
Three checks filter most of the noise. First, does the creator show their own testing, or read the press release back to you? Second, do they link the primary source, whether a paper, a model card or release notes? Third, do they revisit earlier predictions when they turn out wrong? AI Explained, MLST and Latent Space tend to pass all three; most of the loud-thumbnail channels outside this list do not.
Sponsorship is a disclosure question. Many good channels are sponsored, and some, such as 3Blue1Brown, say outright that they take no sponsorships. What matters is whether the sponsored segment is clearly separated from the analysis.
Can YouTube replace AI courses and conferences?
For awareness and much of the technical depth, yes. The AI Engineer channel publishes most conference talks for free within weeks, and Stanford Online posts full lectures that would otherwise sit behind a course fee. What YouTube cannot give you is structure, assessment or the hallway conversations where peers compare vendor contracts and failure stories.
My rule of thumb: use YouTube to decide what to learn, and pay for structure only where a team needs a shared baseline or a credential. A paid course earns its fee when it forces practice on your own systems; a conference earns it through the people you meet. If neither applies, the free channels above cover most of the ground.
How did I choose these 15 channels?
Each channel had to meet four tests: it publishes on AI regularly or has a back catalogue that is still current, it serves technology leaders or engineers rather than general consumers, it has a clear point of view, and its numbers could be checked on the channel page. Subscriber counts and video totals were read on 19 November 2026 and are rounded. Upload frequency comes from the channel's videos tab on the same day.
I left out the official lab channels from OpenAI, Anthropic and Google DeepMind. They are useful for launch keynotes, but they are marketing by design, and the channels above cover the same releases with more scrutiny. I also left out general tech news channels where AI is one topic among many. If you prefer text, the companion hubs for AI newsletters and AI accounts on X apply the same tests, and the AI conferences guide covers the in-person version of AI Engineer.
Listed here, or should be?
If your entry is out of date or you think something belongs on this list, email [email protected]. Entries are editorial and free. A featured, labelled Sponsored slot is available through vendor listings.
Related questions
- Which AI accounts on X are worth following in 2027?
- Which AI newsletters should technology leaders read?
- Which AI conferences are worth attending in 2027?
- What does agentic engineering look like in practice?
- Which agentic AI courses are worth paying for?
- Where should a technology leader start with AI education?
More from AI Education
- AI Education: where technology leaders learn AI in 2027
- AI conferences
- AI accounts on X
- AI newsletters
- AI podcasts for executives
- AI voices on LinkedIn for executives
- AI books for technology leaders
- AI peer communities for executives
- AI courses and university programs
- Agentic engineering resources
- Vibe coding resources
- AI tool stack for CTOs
- The full AI Education hub on We The Flywheel, with niche lists for every role