SiliconvCluster
Dylan Patel - The talent to build your own AI cloud stack does not exist
The analyst who tracks the compute market on what it takes to build an AI cloud, and who is short of the people to do it.
Pyyan / Video
AI on video
Talks, interviews, lectures and podcasts worth the time, from the people who build and argue about AI. Each one carries a line on why it is here, and every link was checked before it was listed. No shorts and no clips: the shortest thing here runs twelve minutes, the longest a little over five hours.
89 videos · newest first · nothing plays until you press it
MoneyThis Week in Startups
Hugging Face Co-Founder on Open-Source, Microduck, and NVIDIA | E2339
Hugging Face's co-founder on open source and Nvidia, recorded while Nvidia's purchase of the company is still closing.
SafetyProf G Markets
He Warned AI Could Destroy Us. Now The Industry Is Listening.
The philosopher who wrote the book on superintelligence twelve years before this argument reached the front pages, on why the industry is only listening now.
SafetyDecoder with Nilay Patel
Microsoft AI CEO says AI threats are real, and Anthropic is making it worse | Decoder
Microsoft's AI chief agrees the threats are real and says Anthropic's way of talking about them makes things worse.
SafetyAlex Kantrowitz
A Sober Conversation About AI Existential Risk — With Nate Soares
Existential risk argued slowly and without the shouting, which is rarer than it should be this month.
PeopleBloomberg Tech
Meta CEO Weighs In on AI Safety Debate
Meta's founder on the safety debate, filling in the one frontier lab that said nothing when the other three called for pacing.
PolicySalesforce
A Conversation with Marc Benioff and H.E. Omar Sultan Al Olama | Dreamforce 2026
Salesforce's founder with the UAE's minister for AI, on what a state does when it decides to run on this.
PolicyAll-In Podcast
JD Vance on AI, Entitlement Fraud, Iran War, Israel, H-1B Abuse & the Midterms
The Vice President on AI, from an administration whose Treasury Secretary has just refused the labs a liability exemption.
ResearchMachine Learning Street Talk
How Physical AI Learns Across Language, Video and Action — Ming-Yu Liu
How models learn across language, video and action, from NVIDIA's lead on physical AI.
AgenticSequoia Capital
Box's Aaron Levie: On Reinventing Yourself in the AI Age and Enterprise Diffusion
What enterprise adoption actually looks like from a chief executive selling into it every day.
MoneyAll-In Podcast
Satya Nadella on the AI Doomer Slowdown, Microsoft’s Master Plan & Who Wins AI
Microsoft's chief executive on the slowdown argument, from the company that owns the largest stake in the lab that started it.
PolicyCBS News
Extended interview: Trump AI adviser David Sacks on tech leaders' AI fears
The White House AI adviser's answer to the chief executives calling for a slowdown.
SiliconAll-In Podcast
Jensen Huang: The Doomer Hoax, Superintelligence is Here, and The Future of AI (ft. President Trump)
Nvidia's chief executive answers the slowdown argument directly, from the company that sells the compute the argument is about.
MoneyBloomberg Tech
AI’s Frontier Faces Calls for Safety Slowdown
Bloomberg on what a frontier slowdown does to the companies whose valuations assume there is not one.
PeopleAll-In Podcast
Elon Musk & Gwynne Shotwell on AI Risks and Peer Review, Starship, Terafab, SpaceX/Tesla Merger
Musk and Gwynne Shotwell on AI risk and peer review, recorded after he agreed in public that Amodei is right.
ResearchMachine Learning Street Talk
Speech Recognition Is Not a Solved Problem — Pavan Muddireddy
Why speech recognition is not finished, in the week Google took the top of the speech to speech board.
SafetyCBS Sunday Morning
Extended Interview: Dario Amodei
The long version of the Amodei interview, days after the essay asking the industry to pace itself.
SafetyFortune Magazine
Altman: AI Beyond Human Control “Absolutely” Possible, Vows Safeguards | Titans and Disruptors
Altman saying AI beyond human control is absolutely possible, in the same week he took a 2026 OpenAI listing off the table.
SafetyCNN
Anthropic CEO reacts to 'AI could kill us all' warning
Amodei on the day he published his essay calling for the industry to pace itself, answering the warning from a researcher who had just left his company.
ResearchMachine Learning Street Talk
What Building an AI Scientist Actually Requires Beyond Intelligence — Edward Hughes
What an AI scientist needs beyond raw intelligence, argued at length rather than demonstrated in a demo.
ResearchDwarkesh Patel
AI researchers debate how close we are to recursive self-improvement
Researchers from inside the labs arguing about how close recursive self improvement is, the question behind this week's calls to slow down.
MoneyBloomberg Podcasts
Why Bridgewater's CIO Says AI's Human Extinction Risk Is Real | Odd Lots
An early backer of both OpenAI and Anthropic, who runs investment at the world's largest hedge fund, on why he thinks extinction risk is real.
AgenticAI Engineer
500 Skills, Zero Fine-Tuning: LinkedIn's Playbook for AI Agents — Ajay Prakash, LinkedIn
How LinkedIn runs five hundred agent skills without fine tuning a model, from the engineer who built it.
ResearchGoogle DeepMind
Can AI help us better predict the weather?
Weather is one of the few places a model's prediction is checked against reality within days, which makes it an honest test.
PeopleNicholas Thompson
The Race to Superintelligence - With Richard Socher
Richard Socher, founder of You.com, interviewed by the chief executive of The Atlantic about the race to superintelligence.
SafetyNBC News
AI researcher says there is 'substantial probability' AI could kill all humans in next decade
Jacob Coxon, the day he resigned from Anthropic, on why he said the people building AI believe it could kill everyone.
SafetyPowerfulJRE
Joe Rogan Experience #2551 - Daniel Kokotajlo
The former OpenAI governance researcher behind the AI 2027 scenario, given two hours on the largest podcast there is.
AgenticArmin Ronacher
State of Agentic Coding #10 with Armin and Ben
Two engineers who use agents daily on what actually changed this month, rather than what was announced.
Researcha16z
Inside OpenAI’s Breakthroughs in Mathematical Reasoning
Two OpenAI mathematicians on how fast AI's mathematical ability is moving, from people who check the proofs rather than market them.
ModelsSources Podcast
Mark Zuckerberg on Muse, Meta's biggest AI bet yet
Zuckerberg on Muse the week Meta put a price on developers' prompts, talking about the agent built on the same model.
AgenticLatent Space
Orbs: Shifting Coding to Cloud — Quinn Slack, Amp Code
What happens to a software company that moves its engineering onto agents, from the chief executive who did it.
ModelsY Combinator
Open Models Change The Economics of AI
Which open models people actually run, from the founder of a tool nine million developers use to run them.
Researcha16z
Why World Models Could Change Robotics, 3D, and Creativity
World models are the line of research that does not run through a chatbot, and this is the clearest recent account of why anyone is betting on them.
PeopleBloomberg Podcasts
Kai-Fu Lee on the Global AI Race | The Mishal Husain Show
Kai-Fu Lee on how Chinese labs are closing the gap under chip restrictions, the week before the US named six of them for distillation.
ModelsBloomberg Television
OpenAI's Sam Altman on Astra Model Debut, Benefits of AI
Altman on the day GPT-6 Astra shipped, to a financial broadcaster rather than a technology podcast.
PeopleTIME
How Close Are We to True Artificial General Intelligence? | OpenAI’s Greg Brockman Speaks to TIME
OpenAI's president on AGI and cybersecurity, recorded the week before Astra shipped and he said it may be the arrival of general intelligence.
SafetyCrowdStrike
Fal.Con 2026: Securing AI
Jensen Huang and Greg Brockman on stage about offensive AI, in the same week a consultancy published the first index of it.
ModelsSources Podcast
Sam Altman on Astra, AGI, and the future of OpenAI
Recorded around the GPT-6 Astra release, with Altman on the AGI claim his own president made rather than a summary of it.
SafetyDwarkesh Patel
Ajeya Cotra: this might be the clearest warning shot we ever get
Two hours on what an early warning of loss of control would actually look like, from someone who has spent a decade on the question.
SafetyDwarkesh Patel
The OpenAI/Hugging Face attack, clearly explained
The July attack is the reason OpenAI delayed a model and added safeguards, and it sits behind the Hugging Face story. Twenty five minutes on what actually happened.
Silicona16z
Why AI Demand Is Outrunning Compute Supply
Gavin Baker on why demand for compute keeps outrunning supply, which is the story behind this week's chip price rises.
PeopleThe Peter McCormack Show
AI CEO: "Your Economic Life Expectancy Ends In 2 Years" | Emad Mostaque
Emad Mostaque at length on work and income. The title is the channel's claim, not a finding; the interview is the argument behind it.
AgenticLex Fridman
DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux | Lex Fridman Podcast #501
Five hours with the creator of Ruby on Rails on what agentic engineering does to programming, from someone with every reason to be sceptical.
FoundationsGoogle DeepMind
The mathematics of AI uncertainty
How a model can know what it does not know. The question behind every confidence field in this site's index.
SiliconDwarkesh Patel
Dylan Patel – Two labs will soon control most of the world's workforce
SemiAnalysis on where the compute actually goes, and the argument that two labs end up holding most of it.
ResearchSequoia Capital
Rich Sutton and Khurram Javed: Why AI Models Stop Learning, and How to Start It Again
Two weeks old. Why models stop learning once training ends, and what a system that kept learning would need.
ResearchAndrew Huberman
Using AI to Increase Your Intelligence and Enrich Humanity
Fei-Fei Li on spatial intelligence to an audience that is not the AI industry, which makes her explain rather than assume.
Researcha16z
Fei-Fei Li is Solving the Hardest Problem in Robotics | World Labs with a16z
World Labs on the robotics problem they think is hardest, which is not the one most people name.
PolicyGZERO Media
Yoshua Bengio: AI is moving faster than our ability to govern it | AI for Good 2026 | Global Stage
Bengio on governance moving slower than capability, recorded before a US senator proposed banning superintelligence outright.
ModelsOpenAI
Introducing ChatGPT Work, powered by Codex and GPT-5.6
OpenAI's own launch stream for ChatGPT Work. Marketing, watched as marketing, and still the primary source.
People20VC with Harry Stebbings
Perplexity CEO: Micron Will Be More Valuable Than Meta & How Export Controls Helped Not Hurt China
Perplexity's founder on export controls and why he thinks a memory company is worth more than Meta.
PeopleBloomberg Live
Thinking Machines’ Murati on AI’s Next Chapter
Murati on what she left OpenAI to build, which she has said very little about anywhere else.
PeopleStanford Graduate School of Business
A Conversation with Demis Hassabis, Co-Founder and CEO of Google DeepMind
An hour with Hassabis in front of a business school audience, which produces different questions from a tech interviewer.
PeopleCNBC International
Europe Wants an AI Champion. Can Mistral Deliver?
Whether Europe can field a frontier lab at all, put to the person running its best attempt.
ResearchUnsupervised Learning: With Jacob Effron
Yann LeCun on What Comes After LLMs
The most senior person publicly arguing that the current approach is a dead end, at length, with the alternative he is building instead.
AgenticSequoia Capital
Andrej Karpathy: From Vibe Coding to Agentic Engineering
Half an hour that named the shift the agentic box on this site is about. The most quoted talk of the year for a reason.
PeopleSequoia Capital
Demis Hassabis: We're Three Quarters of the Way to AGI
A specific claim, with a number attached, from someone in a position to know. Twenty seven minutes.
SiliconLex Fridman
Jensen Huang: NVIDIA - The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494
Huang on how the company selling the shovels came to be worth four trillion dollars.
PeopleStarTalk
Is AI Hiding Its Full Power? With Geoffrey Hinton
Hinton with an interviewer who pushes back, which is rarer than it should be. Three million views.
PeopleDwarkesh Patel
Dario Amodei — “We are near the end of the exponential”
The interview where he said the exponential is ending. Over a million views and quoted all year.
PeopleFinancial Times
Mustafa Suleyman sets out Microsoft AI's goal of 'humanist superintelligence' | FT Interview
Microsoft AI's stated goal, in his own words, to an interviewer who asks what the words mean.
SiliconLex Fridman
State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI | Lex Fridman Podcast #490
A four hour state of the field: scaling, China, agents, GPUs. The single best catch-up if you have been away.
ResearchGoogle DeepMind
The future of intelligence, with Demis Hassabis
Hassabis on his own lab's channel with Hannah Fry asking real questions. Ten million views, and it earns them.
ResearchGoogle DeepMind
AlphaFold: Grand challenge to Nobel Prize | John Jumper
John Jumper on AlphaFold, from the problem to the Nobel Prize, on DeepMind's own channel.
ResearchDwarkesh Patel
Ilya Sutskever: we're moving from the age of scaling to the age of research
He says almost nothing in public. This is the interview where he argued scaling alone has stopped being the answer, before the field agreed.
ResearchDwarkesh Patel
Richard Sutton – Father of RL thinks LLMs are a dead end
The man who wrote the book on reinforcement learning arguing that language models are a dead end.
ResearchAnthropic
Interpretability: Understanding how AI models think
Anthropic's own researchers on looking inside a model. The best public introduction to interpretability.
ResearchY Combinator
Scaling and the Road to Human-Level AI | Anthropic Co-founder Jared Kaplan
Scaling laws from the person who wrote them down, which is a different talk from anyone else giving it.
PeopleThe Royal Institution
Will AI outsmart human intelligence? - with 'Godfather of AI' Geoffrey Hinton
The Royal Institution lecture. Same argument, made to a room rather than a camera, and better for it.
PeopleY Combinator
Elon Musk: Digital Superintelligence, Multiplanetary Life, How to Be Useful
Musk to a room of founders rather than a broadcaster, which changes what he bothers to explain.
SafetyTED
The Catastrophic Risks of AI — and a Safer Path | Yoshua Bengio | TED
Fifteen minutes. If you want the risk case from a Turing laureate without an afternoon to spare, this is it.
ResearchSequoia Capital
Google's Jeff Dean on the Coming Transformations in AI
The person who built much of Google's infrastructure on what he thinks changes next.
FoundationsStanford Online
Stanford CS336 Language Modeling from Scratch | Spring 2025 | Lecture 1: Overview and Tokenization
Stanford's language modelling course, taught properly over a term. The 2026 lectures have embedding switched off, so this is the edition you can watch here.
BenchmarksARC Prize
ARC-AGI-2 Overview With Francois Chollet
The ARC-AGI-2 rules, explained by the person who set them. Twenty minutes, and it changes how you read a benchmark score.
ResearchVeritasium
AlphaFold - The Most Useful Thing AI Has Ever Done
Ten million people have watched this account of the most useful thing AI has done. It deserves them.
FoundationsAndrej Karpathy
Deep Dive into LLMs like ChatGPT
Three and a half hours from tokenisation to RLHF, aimed at people who want the whole machine rather than an analogy.
SiliconCNBC
How China’s New AI Model DeepSeek Is Threatening U.S. Dominance
Six million views on how a Chinese lab rearranged everyone's assumptions. Useful background to this year's open weights run.
ResearchMachine Learning Street Talk
François Chollet on OpenAI o-models and ARC
Chollet on reasoning models and why his benchmark keeps being harder than it looks.
PeopleLex Fridman
Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast #452
Five hours with Anthropic's chief executive. Long enough that the media-trained answers run out.
Foundations3Blue1Brown
How might LLMs store facts | Deep Learning Chapter 7
Where a fact physically lives inside a model, which turns out to be a real question with a partial answer.
ResearchAnthropic
Scaling interpretability
The follow-up: what breaks when you try to do interpretability on a frontier-sized model.
ResearchTED
With Spatial Intelligence, AI Will Understand the Real World | Fei-Fei Li | TED
The TED talk that put spatial intelligence in the vocabulary. Fifteen minutes and still the clearest statement of it.
FoundationsTED
What Is an AI Anyway? | Mustafa Suleyman | TED
The definition question, asked properly. Nearly three million people have watched someone try to say what an AI is.
Foundations3Blue1Brown
Attention in transformers, step-by-step | Deep Learning Chapter 6
Attention, drawn rather than described. The follow-up to the transformers chapter and the part most explanations skip.
Foundations3Blue1Brown
Transformers, the tech behind LLMs | Deep Learning Chapter 5
The clearest explanation of a transformer that exists, and eleven million people agree. Start here if the architecture has never quite landed.
FoundationsAndrej Karpathy
[1hr Talk] Intro to Large Language Models
The hour that taught a lot of the industry how to talk about LLMs. Older now, and still the cleanest overview.
SafetyCITRIS and the Banatao Institute
How Not To Destroy the World With AI - Stuart Russell
The control problem, laid out by the person who wrote the textbook most of the field learned from.
FoundationsAndrej Karpathy
The spelled-out intro to neural networks and backpropagation: building micrograd
Backpropagation built from nothing, in Python, with every line explained. The one to watch if you learn by making it.
Foundations3Blue1Brown
But what is a neural network? | Deep learning chapter 1
Twenty four million views and still the best first hour anyone has spent on neural networks.
Featuring Clément Delangue, Nick Bostrom, Mustafa Suleyman, Mark Zuckerberg, Satya Nadella, Jensen Huang, Elon Musk, Dario Amodei, Sam Altman, Kai-Fu Lee, Greg Brockman, Emad Mostaque, Richard Sutton, Fei-Fei Li, Yoshua Bengio, Aravind Srinivas, Mira Murati, Demis Hassabis, Arthur Mensch, Yann LeCun, Andrej Karpathy, Geoffrey Hinton, Ilya Sutskever, Jared Kaplan, Jeff Dean, Stuart Russell and others from the index of people.