YouTube · AI
Top AI Creators to Follow in 2026
22 of the most-followed YouTube channels that explain AI in 2026: research breakdowns, news, interviews, coding and everyday work, with a video each.
Short answer
The top AI creators on YouTube in 2026 are the channels that explain the research, test the models, interview the people building them, and show how AI fits into real work. This list has 22 of them, grouped by what they do: research and papers explained (6), news and analysis (4), interviews and podcasts (5), coding and building with AI (5), and AI for everyday work (2). Every entry links one recent video to start with.
This list is for anyone who wants to understand AI rather than be shouted at about it: a paper walked through, a new model tested on something real, a researcher given two hours to think out loud, or a plain tutorial on using an assistant at work. With one exception the entry explains, the channels below are among the most-followed in the niche, and all have kept uploading through the last three months.
“Top” here means the most-followed channels in the niche that are still active and pass our checks, not an order. The entries are alphabetical, because a channel deriving activation functions on a whiteboard and a daily news show serve different viewers, and putting one above the other would say nothing useful. If you are choosing AI tools rather than trying to understand AI, the sibling list of AI tool reviewers on YouTube is the one you want; nobody appears on both.
Which AI creators should you follow?
Start with what you want from them. Everyone in the table is on the list below, and within each row they’re alphabetical.
| If you want… | Creators to start with | What they do |
|---|---|---|
| The research explained | 3Blue1Brown, AI Explained, bycloud, StatQuest with Josh Starmer, Two Minute Papers, Welch Labs | Papers, model cards and the maths underneath them, walked through with the sources linked |
| News and analysis | Matt Wolfe, Matthew Berman, Nate B Jones (AI News & Strategy Daily), The AI Daily Brief | Weekly and daily roundups with primary links, and new models put to work rather than just announced |
| Interviews and podcasts | Dwarkesh Patel, Latent Space, Lex Fridman, Machine Learning Street Talk, No Priors | Long conversations with researchers, lab founders and engineers, with transcripts or chapter lists |
| Coding and building with AI | Dave Ebbelaar, IndyDevDan, Prompt Engineering, Sam Witteveen, sentdex | Agents, fine-tuning, benchmarks and home hardware, built on screen with the code published |
| AI for everyday work | Kevin Stratvert, Tina Huang | Step-by-step tutorials on assistants, coding agents and local models, prompts included |
New to all of this? Start with the everyday-work row. Already building with models? The coding row and the two technical podcasts will feel more like home.
The 22 creators, A to Z
3Blue1Brown
- Covers
- Visual maths explainers, including the AI ones
- Why they’re here
- Grant Sanderson explains maths and adjacent computer science by animating the core ideas, per his About text. The video to start with takes the one Olympiad problem that eluded AI, builds the best construction, proves it optimal, and ends with reflections on AI in maths. A maths channel first; the AI episodes come now and then.
- Start with
- The last IMO problem AI could not solve
Is this you? Update or claim · Remove me
AI Explained
- Covers
- Model releases and lab claims, read closely
- Why they’re here
- Reads the primary documents and links them. The video to start with digs into the 230-page paper behind Claude Opus 5.5, asks whether the labs’ goalposts for automated AI research keep moving and whether these models can be tested at all, with the system card and three versions of Anthropic’s scaling policy linked. He also runs Simple Bench, his own benchmark.
Is this you? Update or claim · Remove me
AI News & Strategy Daily | Nate B Jones
- Covers
- Daily AI strategy takes for people at work
- Why they’re here
- Daily videos and weekly strategy breakdowns from a product leader, with a written version on his newsletter. The video to start with argues that a leaderboard win says little about usefulness: Gemini 4 Argon tops the Vals Index while almost nobody can use it yet, and running his own subscriptions through two rival agents shows that the product experience, not the model, is the bottleneck.
Is this you? Update or claim · Remove me
bycloud
- Covers
- Frontier AI research papers, explained
- Why they’re here
- “Frontier AI Research Explained” is the whole About text, and the videos keep to it. The video to start with breaks down how Jev, a decision model that doesn’t generate words, works under the hood, where it came from and what the alternatives are, with the papers it draws on linked. The sponsor gets YouTube’s paid-promotion label and the description’s first line.
Is this you? Update or claim · Remove me
- Covers
- Production AI systems built on camera
- Why they’re here
- Builds production systems on camera rather than demos. The video to start with turns local Markdown files into a company knowledge base, moves it into a CMS so teammates can edit it, compares keyword, semantic and hybrid search, and adds a chat widget that answers with links to the source pages, with the code on GitHub. It carries YouTube’s paid-promotion label; the partner’s link is in the description.
Is this you? Update or claim · Remove me
- Covers
- Long interviews with AI researchers
- Why they’re here
- Host of the Dwarkesh Podcast, a long-form interview show on AI, science and history, so AI episodes alternate with others. The episode to start with sits down with OpenAI researcher Noam Brown on multi-agent systems, what the recent run of maths results says about automating AI research, and how anyone would know alignment is solved before that starts; a transcript is linked. Sponsors get their own heading and YouTube’s label.
Is this you? Update or claim · Remove me
- Covers
- Agentic engineering, benchmarks and model choice
- Why they’re here
- An engineer who builds with coding agents and shows the decisions behind it; “avoid hype” and “build real products” are the channel’s stated rules. The video to start with picks the five benchmarks he actually uses, ranks GPT-6 Astra, Claude Fable 5.1 and the open-weight models on each, and argues that a model choice weighs performance, cost and speed together, not one index score. He sells his own courses, linked from each video.
Is this you? Update or claim · Remove me
Kevin Stratvert
- Covers
- Step-by-step AI tool tutorials for beginners
- Why they’re here
- A former Microsoft product manager who teaches software step by step, with AI tools now a large part of the schedule. The video to start with builds a small browser game from one prompt in OpenAI’s Codex, then shows previewing, tracking changes with Git, running agents in parallel and handing work to the cloud, with the exact prompt pasted in the description, which calls the tutorial beginner-friendly.
Is this you? Update or claim · Remove me
- Covers
- The AI engineers’ podcast and newsletter
- Why they’re here
- A podcast and newsletter for people who build with models, with full show notes on its site. The episode to start with is a panel with Google DeepMind’s Pushmeet Kohli and Biohub’s Sal Candido on why AlphaFold was only the beginning for protein folding, why scaling compute and data alone won’t solve biology, and what a virtual cell would need. Technical, and better for it.
Is this you? Update or claim · Remove me
- Covers
- Long-form conversations, AI among them
- Why they’re here
- A computer scientist whose podcast ranges across science, technology, sports and politics, so AI episodes arrive between others. The one to start with is five hours with Ruby on Rails creator DHH on programming with AI agents, vibe coding versus agentic engineering, the best coding models and harnesses, and what happens to open source, with a transcript and timestamped outline linked. Sponsors are listed by name in the description.
Is this you? Update or claim · Remove me
- Covers
- Technical research interviews and philosophy of AI
- Why they’re here
- Long, technical interviews with researchers; the About text describes the show as covering advanced AI research, science and philosophy. The episode to start with has Prior Labs co-founder Frank Hutter on TabPFN, a model for tables that is pre-trained on synthetic data and predicts in a single forward pass, why deep learning struggled with tables for a decade, and how the work grew out of AutoML, with a table of contents.
Is this you? Update or claim · Remove me
Matt Wolfe
- Covers
- Weekly AI news roundups with sources
- Why they’re here
- A Friday news roundup, per the About text, which points to the FutureTools site and newsletter for more. The video to start with covers one week’s releases, from OpenAI’s Dots and GPT-6.1 Sol to Claude Sonnet 5.5 and Gemini 4 Argon, with a primary link for every story in the description, mostly the labs’ own announcements. The sponsor is named in the first lines and the video carries YouTube’s label.
Is this you? Update or claim · Remove me
Matthew Berman
- Covers
- New models tested by building things
- Why they’re here
- Daily coverage of new models, often by building with them. The video to start with reacts to Claude Sonnet 5.5 with five projects built on it, Crownfall and Bounce Lab among them, each published at its own link so you can open the result, plus two comparison builds from the previous Sonnet. The sponsor is named at the top of the description and the video carries YouTube’s paid-promotion label.
- Start with
- How Did Anthropic Pull This Off?
Is this you? Update or claim · Remove me
- Covers
- Investors interviewing AI founders and researchers
- Why they’re here
- Investors Sarah Guo and Elad Gil interview engineers, researchers and founders, per the show’s About text. The episode to start with has ReflectionAI’s Misha Laskin on Beam, the company’s open-weight reasoning model: the pre-training and reinforcement learning behind it, whether open models will take most token demand, and the safety questions open weights raise. Smaller on YouTube than most here, but a weekly conversation with the people building the models.
Is this you? Update or claim · Remove me
- Covers
- Fine-tuning, RAG and agents, hands-on
- Why they’re here
- Hands-on tutorials from an ML lead with a PhD who, per his About text, shares what he learns without the “fluff” and “hype”. The video to start with fine-tunes Google’s Embedding Gemma 2 in a free Colab notebook on two tasks, reports the before-and-after accuracy of each, checks what the tuning costs the rest of the model, and flags a reload bug that silently breaks audio adapters.
Is this you? Update or claim · Remove me
- Covers
- Agent design and open models, on screen
- Why they’re here
- Builds agents on screen and explains the design choices; per his About text he has worked with transformers and LLMs since 2018. The video to start with wires a small decision model into an agent loop as a “smart if statement” for routing, risk gating and tool selection, then gives a chapter to where it doesn’t fit, including prompt injection, with two live demos and the code on GitHub.
- Start with
- Using Jev In Your Agent Harness
Is this you? Update or claim · Remove me
- Covers
- Python, robots and frontier AI at home
- Why they’re here
- A long-running Python channel that now spends much of its time trying to run frontier AI at home. The video to start with continues that series, covering the PCIe switching behind his build and how DeepSeek V4 Flash, GLM 5.2, Qwen 3.8 and others behave on it, with the benchmarks written up on his site and the harness on GitHub. The hardware he bought is linked.
Is this you? Update or claim · Remove me
- Covers
- Statistics and machine learning, step by step
- Why they’re here
- Josh Starmer teaches statistics and machine learning as small, simple steps and, per his About text, doesn’t dumb the material down. The video to start with opens with a “too long, didn’t watch” summary, then derives the GELU, SiLU and SwiGLU activation functions used in today’s models, starting from why sigmoid and ReLU fail. No sponsor on that video; he sells books and study guides from his site.
Is this you? Update or claim · Remove me
The AI Daily Brief: Artificial Intelligence News
- Covers
- Daily AI news and discussion
- Why they’re here
- A daily news show, also published as a podcast and a newsletter. The episode to start with is a practical one: with guest Nufar Gaspar, the host builds a repeatable way to test new models against your own tasks, comparing outputs on quality, speed and cost before deciding whether to switch; the show’s own site calls it a webinar edition. An ad-free version is offered through Patreon.
- Start with
- The Best Way To Test New AI Models
Is this you? Update or claim · Remove me
- Covers
- AI, coding and local models made approachable
- Why they’re here
- An ex-Meta data scientist, per her About text, who makes AI approachable without skipping the hardware. The video to start with runs local models of every size on every machine she could find, from a Raspberry Pi and a phone up to a Mac Studio and an eight-GPU server, with a hardware lesson and two quizzes. The sponsor’s offer is the description’s first line; affiliate links sit under an “Affiliates” heading.
- Start with
- Every Size Local AI In 24 Minutes
Is this you? Update or claim · Remove me
- Covers
- New AI papers in short walkthroughs
- Why they’re here
- Short paper walkthroughs, about twice a week, each built around one published result. The video to start with walks through DeepSeek’s OCR paper, linked in the description; the About text is just the catchphrase “What a time to be alive!”. Titles run hot, but the source is always there. The sponsor gets the first line of the description and YouTube’s paid-promotion label.
- Start with
- DeepSeek Solved AI’s Memory Problem
Is this you? Update or claim · Remove me
- Covers
- Illustrated deep dives on AI foundations
- Why they’re here
- Long, illustrated explainers of the ideas underneath modern AI; the About text points to the team’s illustrated guide to AI. The video to start with explains residual networks, the most cited paper of this century, from why deep networks stopped improving through loss landscapes to the residual stream inside today’s transformers, with references listed and a note that it rests on the team’s papers and talks. The sponsor gets YouTube’s label and a chapter.
Is this you? Update or claim · Remove me
How we checked each creator
We looked at about 70 channels across AI research explainers, news, podcasts, coding and productivity, found through 27 YouTube searches, creators’ own recommendations, and the usual “best AI channels” round-ups (used to find names, never as a source of facts). For each one that made the list we:
- Checked their reach. We set a subscriber floor for this list and read each channel’s public count in October 2026. The floor exists; the numbers aren’t printed, for the reasons in the FAQ.
- Checked they’re active. Every creator here uploaded at least three times in the last 90 days, confirmed against their channel’s public upload feed. Two famous channels that explain AI very well didn’t make it because their feeds showed nothing in that window.
- Picked one real video. A recent upload where they explain, test or discuss AI on screen, with its exact title and channel confirmed by YouTube itself. For three creators we also took one sentence about who they are from English Wikipedia, and nothing else.
- Read what the video sets out to do. Each “why” comes from that video’s own description and chapters and the channel’s About text: what it covers, what is linked, and how any sponsor or affiliate link is disclosed. Where a famous channel’s descriptions gave us nothing to say, we picked another video rather than paraphrasing a reputation.
- Looked for the warning signs. We left out channels built on hype titles, mass-produced AI-generated uploads or “make money with AI” pitches, however large, along with corporate channels, which aren’t creators, and broad programming or tech channels that cover AI as one topic among many.
We did not read the comments, so this list says nothing about the conversation under each video; that is a job for the first quarterly review. Our methodology covers the rules every list follows.
How to follow AI channels without getting sold to
Most channels here take sponsors, run affiliate links or sell a course, and several do all three. None of that makes them worse to watch, as long as you can see it. Four habits:
- Look for YouTube’s paid-promotion label at the start. When a creator declares branded content, YouTube adds a disclosure label that appears at the beginning of the video. Nine of the 22 videos we picked here carry it or name a sponsor in the description’s first lines. If a video reads like an ad and has neither, be careful.
- Don’t accept a disclosure you have to dig for. The FTC’s endorsement guides say that “consumers can easily miss disclosures in the video description” and that a disclosure behind a “more” link is “easily avoidable”. The good channels here say it out loud and in the first line.
- Know what an affiliate link is. The FTC’s view is that “paid link” right next to an affiliate link is an adequate disclosure, and that the same guidance “applies anytime you endorse a product and get paid through affiliate links”. Some creators above label affiliate links under their own heading; others link hardware they bought with no label at all. Our guide to disclosing affiliate links and sponsorships covers what each platform expects.
- Check the AI-content label on anything that looks too polished. YouTube requires creators to disclose when they use AI to make a real person appear to say or do something they didn’t, alter footage of a real event, or generate a realistic scene; the label appears in the expanded description or on the player. It won’t catch everything, but a “news” channel with a synthetic presenter and that label is telling you what it is.
When the video is a review of a specific tool, the same logic applies with a few more checks; we go through them in how to tell if a YouTube AI tool review is honest.
What to do next
- Want to understand the field? Pick one channel from the research row and one from the interviews row, and give each a full video before deciding.
- Choosing AI tools? Go to the AI tool reviewers list, which is built for that question.
- A brand looking for AI creators? Read how to vet a YouTube creator before a sponsorship first; most of the channels here list a business contact, and the sponsored videos we picked disclose it.
- Know someone who belongs here? Suggest a creator. Suggestions go through the same checks as everyone else, and we review this list every quarter.
Frequently asked questions
- Who are the best AI creators to follow on YouTube in 2026?
- It depends on what you want from them. To understand the research: 3Blue1Brown, AI Explained, bycloud, StatQuest with Josh Starmer, Two Minute Papers and Welch Labs. For news and analysis: Matt Wolfe, Matthew Berman, Nate B Jones and The AI Daily Brief. For long interviews with researchers and builders: Dwarkesh Patel, Latent Space, Lex Fridman, Machine Learning Street Talk and No Priors. For coding and building with AI: Dave Ebbelaar, IndyDevDan, Prompt Engineering, Sam Witteveen and sentdex. For AI in everyday work: Kevin Stratvert and Tina Huang.
- Did any of these creators pay to be on this list?
- No. Nobody pays to be included, inclusion never depends on using or promoting any product, and nobody on this list was asked to mention one. Each creator is here because of their reach, their verified activity and one recent video in which they explain, test or discuss AI on screen.
- How did you choose these AI creators?
- We started with about 70 channels found through YouTube searches and creators’ own recommendations, kept the ones above a subscriber floor we set for this niche, confirmed at least three uploads in the last 90 days from each channel’s public feed, then picked one recent video per creator and verified its title and channel with YouTube itself. Hype-title, mass-produced and “make money with AI” channels were left out, however large. Our methodology page has the details.
- What does “top” mean here, and why is the list alphabetical?
- “Top” means the most-followed channels in the niche that are still active and pass our checks. It is not an order: a channel deriving activation functions and a daily news show serve different viewers, and ranking them against each other would say nothing useful. Alphabetical order also makes it obvious that no position was bought or traded.
- Why are there no subscriber counts on this page?
- We only publish YouTube numbers that come from the official YouTube Data API, dated, and refreshed or deleted within 30 days, as YouTube’s developer policies require. We read each channel’s public count when choosing the list, but a scraped number isn’t one we’ll print; until API numbers are in place we’d rather show none.
- Where are the AI tool reviewers?
- On their own list. This page is for channels that explain AI, report on it, interview the people building it and show how they use it. Creators who test specific tools on screen are on our AI tool reviewers list, and nobody appears on both.
About this list
We review every list each quarter. This one was last reviewed on . Every number is dated and says where it came from. How we pick creators.
Spot a mistake? Email corrections@topfluencers.co and we’ll fix it within 48 hours.
Featured here? There’s a free badge you’re welcome to use, with no conditions: get your badge.
More lists
- YouTube · AI tools
AI Tool Reviewers on YouTube Worth Following in 2026
21 YouTube creators who test AI tools on camera, show the real output and say when something doesn’t work. Each entry says why, with a video to start with.
21 creators · Reviewed
Read the list - YouTube · Shopify and ecommerce
Shopify Creators Who Teach Running a Real Store
21 creators who teach Shopify from the operator’s side: store setup, ads with real numbers, discounts, wholesale and theme code. No get-rich-quick pitches.
21 creators · Reviewed
Read the list