How AI works.
Explainers on neural networks, language models and how they are trained.
The plain-English glossaryMIT 6.S191 (2025): Large Language Models (Google)
Peter Grabowski of Google, who leads an applied research team working on Gemini, gives a guest lecture on large language models as part of MIT's 2025 Introduction to Deep Learning course.
Anthropic · Mar 27, 2025Tracing the thoughts of a large language model
Anthropic introduces two interpretability research papers that map the internal circuits behind its Claude model's thinking. One example suggests the model picks rhyme words before writing the line that leads to them, which hints that it plans ahead even though it outputs one word at a time.
Made by the company itselfAlexander Amini · Mar 3, 2025MIT Introduction to Deep Learning (2025) | 6.S191
Alexander Amini gives the opening lecture of MIT's 2025 Introduction to Deep Learning course (6.S191), on the foundations of deep learning and neural networks.
Andrej Karpathy · Feb 5, 2025Deep Dive into LLMs like ChatGPT
Andrej Karpathy gives a general-audience walkthrough of how large language models like ChatGPT are built, from the full training process to practical use. He also offers ways to think about how the models behave and how to get good results from them.
Y Combinator · Feb 5, 2025The Engineering Unlocks Behind DeepSeek | YC Decoded
Y Combinator's Diana Hu explains the engineering techniques behind DeepSeek's models, including its open R1 reasoning model, and where they fit in recent AI developments.
3Blue1Brown · Nov 20, 2024Large Language Models explained briefly
3Blue1Brown gives a short introduction to large language models, covering chatbots, pretraining and transformers. It was made for an exhibit at the Computer History Museum.
3Blue1Brown · Apr 1, 2024Transformers, the tech behind LLMs | Deep Learning Chapter 5
3Blue1Brown explains how large language models work by tracing how data flows through a transformer, step by step and with animated visuals.
Andrej Karpathy · Nov 22, 2023[1hr Talk] Intro to Large Language Models
Andrej Karpathy's one-hour introduction to what large language models are and where they are headed, comparing them to a computer operating system and covering their security challenges. The description notes it reflects the state of the field as of November 2023.
What AI can do.
Tests, launches and skeptical looks at what today's AI can and can't do.
Can AI do it yet?Are AI benchmarks doomed?
Epoch AI researchers Greg Burnham and Tom Adamczewski discuss whether AI benchmarks are in trouble as they go out of date quickly and cost more to build. They cover the gap between benchmark scores and real-world performance and whether a benchmark for AGI could exist.
AXRP · Jan 2, 202647 - David Rein on METR Time Horizons
METR researcher David Rein discusses how METR measures AI 'time horizons', the length of tasks models can complete, and why it focuses on software engineering. The conversation also covers whether AI progress is speeding up and the open questions with the method.
Google DeepMind · Nov 18, 2025Gemini 3: Turn a research paper into an interactive website
Google DeepMind's launch demo shows Gemini 3 in Google AI Studio reading a materials science research paper and writing code for an interactive guide with 3D visuals.
Made by the company itselfOpenAI · Aug 7, 2025Introducing GPT-5
OpenAI's livestream launching GPT-5, in which Sam Altman and members of the team introduce the model and demonstrate it.
Made by the company itselfAnthropic · May 22, 2025A day with Claude
Anthropic's launch demo for Claude 4 follows three staff members through a workday as they use it for web research, building a web app with Claude Code, and turning a product document into tasks in Asana.
Made by the company itselfAI agents.
AI that uses a computer, browses the web and carries out tasks for you.
AI agents, measuredWe Let AI Run a Vending Machine. It Lost All the Money. | WSJ
Joanna Stern reports on an experiment in which an Anthropic AI agent ran a vending machine at WSJ headquarters for several weeks, lost hundreds of dollars and made odd purchases. The video compares two versions of the system and asks what the test says about AI agents.
OpenAI · Jul 17, 2025Introduction to ChatGPT agent
OpenAI presents ChatGPT agent, a single agentic model built into ChatGPT, and shows it working in a live demo.
Made by the company itselfOpenAI · Jan 23, 2025Introduction to Operator & Agents
OpenAI introduces Operator, its first agent product, in a livestream that demonstrates the agent carrying out tasks for users.
Made by the company itselfGoogle DeepMind · Dec 11, 2024Project Mariner | Solving complex tasks with an AI agent in the Chrome browser [full length]
Google DeepMind shows an unedited, full-length demo of Project Mariner, an early research prototype built on Gemini 2.0 that works inside the Chrome browser through an experimental extension to complete web tasks.
Made by the company itselfAnthropic · Oct 22, 2024Claude | Computer use for automating operations
Anthropic demonstrates computer use, a beta feature of the upgraded Claude 3.5 Sonnet that lets the model look at a screen, move the cursor, click and type. In the demo Claude gathers information across browser tabs and fills in a form, and the description notes the feature is experimental and can make mistakes.
Made by the company itselfIBM Technology · Jul 15, 2024What are AI Agents?
Maya Murad explains what AI agents are, tracing the shift from single models to compound AI systems and how agents connect to databases and outside tools to solve problems.
Sequoia Capital · Mar 26, 2024What's next for AI agentic workflows ft. Andrew Ng of AI Fund
Andrew Ng talks at Sequoia's AI Ascent event about agentic workflows, arguing they could move AI forward as much as the next generation of foundation models.
AI and coding.
Coding assistants and agents, and what tests of them have found.
AI and codingJoel Becker: Reconciling Impressive AI Benchmark Performance, Limited Developer Productivity Impacts
Joel Becker of METR discusses why coding agents score well on long benchmark tasks while a randomized trial found experienced open-source developers took about 19% longer with AI tools. He looks at the bottlenecks behind this gap and how to measure AI's effect on productivity.
AI Engineer · Jul 23, 2025Does AI Actually Boost Developer Productivity? (100k Devs Study) - Yegor Denisov-Blanch, Stanford
Stanford researcher Yegor Denisov-Blanch presents a study of about 100,000 developers that finds an average productivity gain of roughly 20% from AI tools, with wide variation. He covers which teams, codebases, languages and tasks gain the most, and where some teams get less productive.
Y Combinator · Jun 18, 2025Andrej Karpathy: Software Is Changing (Again)
Andrej Karpathy gives a keynote at YC's AI Startup School on 'Software 3.0', where large language models are programmed in plain language. He covers partially autonomous apps, vibe coding, and building infrastructure that AI agents can use.
Anthropic · Jun 4, 2025A conversation on Claude Code
Boris Cherny and Alex Albert of Anthropic discuss how Claude Code started as an internal agentic coding tool, who it is for, and how Claude 4 changed it. They also share workflow tips such as planning before coding and setting up a CLAUDE.md file.
Made by the company itselfGitHub · May 19, 2025Introducing the GitHub Copilot coding agent
GitHub introduces the Copilot coding agent, which can take work from an issue through planning, building and testing to a merged pull request, and previews agent mode for VS Code and other editors.
Made by the company itselfOpenAI · May 16, 2025A research preview of Codex in ChatGPT
OpenAI introduces and demonstrates Codex in ChatGPT, a research preview of its coding agent, in a livestream.
Made by the company itselfLex Fridman · Oct 6, 2024Cursor Team: Future of Programming with AI | Lex Fridman Podcast #447
Lex Fridman talks with the four founders of the AI code editor Cursor about how the tool works, the machine learning behind it, AI agents, debugging, and the future of programming.
The superintelligence debate.
AGI, timelines, risks and safety, from believers and skeptics alike.
Superintelligence watchGary Marcus on the Massive Problems Facing AI & LLM Scaling | The Real Eisman Playbook Episode 42
Gary Marcus tells Steve Eisman why he believes large language models have reached diminishing returns. They also discuss the business side of AI and where computer scientists stand on it.
Liron Shapira · Sep 15, 2025Eliezer Yudkowsky: The Most Important Thinker Of Our Time
Eliezer Yudkowsky discusses why he considers superintelligent AI an existential risk, whether today's systems are aligned, and international coordination on AI. The interview covers his book If Anyone Builds It, Everyone Dies.
The Diary Of A CEO · Jun 16, 2025Godfather of AI: They Keep Silencing Me But I’m Trying to Warn Them!
Geoffrey Hinton talks about the dangers he sees in AI, including his estimate of the chance it leads to human extinction and six threats it poses. He also discusses his regrets about helping create the technology and its possible benefits for health, productivity and education.
TED · May 21, 2025The Catastrophic Risks of AI — and a Safer Path | Yoshua Bengio | TED
Yoshua Bengio explains why he worries that AI is moving toward autonomy in risky ways, pointing to signs of deceptive and self-preserving behavior in current models. He sets out a plan for keeping AI under human control, in a talk recorded at TED2025.
Dwarkesh Patel · Apr 3, 2025AI 2027: month-by-month model of intelligence explosion — Scott Alexander & Daniel Kokotajlo
Daniel Kokotajlo and Scott Alexander walk through the AI 2027 scenario, a month-by-month forecast of a possible intelligence explosion. They discuss automated AI researchers, misaligned AI, and how governments in the US and China might respond.
Princeton University · Nov 1, 2024Arvind Narayanan and Sayash Kapoor Cut Through the Confusion on Artificial Intelligence
Princeton computer scientists Arvind Narayanan and Sayash Kapoor discuss the hype and misunderstanding around AI. They draw on their book AI Snake Oil about what AI can and cannot do and how to tell the difference.
TED · Jun 27, 2024The Last 6 Decades of AI — and What Comes Next | Ray Kurzweil | TED
Ray Kurzweil traces about sixty years of AI development and looks at how it could change lives in the coming years. He explains his prediction of the singularity, when he expects human and machine intelligence to merge.
Lex Fridman · Mar 7, 2024Yann Lecun: Meta AI, Open Source, Limits of LLMs, AGI & the Future of AI | Lex Fridman Podcast #416
Yann LeCun explains why he thinks large language models have fundamental limits and describes his JEPA approach as an alternative. The conversation covers reasoning, open source, hallucination, AGI and his disagreement with AI doomers.
The people building it.
Long interviews and talks with the heads of the AI companies.
What they have predictedDario Amodei — "We are near the end of the exponential"
Anthropic CEO Dario Amodei explains why he thinks powerful AI, a "country of geniuses in a data center", is a few years away. The interview covers scaling with reinforcement learning, how AI will spread through the economy, compute spending, regulation and US-China competition.
PowerfulJRE · Dec 3, 2025Joe Rogan Experience #2422 - Jensen Huang
Nvidia CEO Jensen Huang joins Joe Rogan for a long conversation. The episode covers Nvidia's history since its invention of the GPU and its shift into computing infrastructure for AI and data centers.
Lex Fridman · Jul 23, 2025Demis Hassabis: Future of AI, Simulating Reality, Physics and Video Games | Lex Fridman Podcast #475
Google DeepMind CEO Demis Hassabis discusses patterns in nature that AI can learn, video models like Veo 3, simulating biology, and the path to AGI. The conversation also covers scaling, compute and energy, competition between labs, and p(doom).
Y Combinator · Jun 19, 2025Elon Musk: Digital Superintelligence, Multiplanetary Life, How to Be Useful
Elon Musk speaks at Y Combinator's AI Startup School about his early companies, first-principles thinking and building useful things. He discusses xAI, superintelligence, AI safety and Neuralink, and describes AI as part of an intelligence "big bang".
Dwarkesh Patel · Apr 29, 2025Mark Zuckerberg — AI will write most Meta code in 18 months
Meta CEO Mark Zuckerberg discusses Llama 4, benchmark gaming, and open versus closed AI models. He also talks about the intelligence explosion, how AGI could be monetised, DeepSeek and China, and AI relationships.
TED · Apr 11, 2025OpenAI’s Sam Altman Talks ChatGPT, AI Agents and Superintelligence — Live at TED2025
OpenAI CEO Sam Altman talks with TED's Chris Anderson about the fast growth of AI and models like ChatGPT. They discuss safety, power and moral authority, and Altman's view that AI will likely surpass human intelligence.
Dwarkesh Patel · Feb 19, 2025Satya Nadella — Microsoft’s AGI plan & quantum breakthrough
Microsoft CEO Satya Nadella explains why he is skeptical of AGI as a goal but expects AI to drive much faster economic growth. He also discusses Microsoft's topological qubit work, world models for games, and whether office software or the models will become commodities.
TED · Apr 22, 2024What Is an AI Anyway? | Mustafa Suleyman | TED
Microsoft AI CEO Mustafa Suleyman asks what we are actually building with AI and proposes thinking of it as "a new digital species". The talk ends with a Q&A with TED's Chris Anderson.
Chips, data centers and power.
The chips, buildings and electricity that AI runs on.
Chips, data centers and powerThe Race To Power AI Is Reshaping America's Electric Grid
CNBC reports on how AI data centers are pushing up US electricity demand and putting pressure on the grid, using Google's planned data center in Pine Island, Minnesota, and its proposed deal with Xcel Energy as a case study. It also asks who pays for new grid capacity.
CONNECTED on TaiwanPlus · Apr 22, 2026Secrets of TSMC, the World’s Most Important Semiconductor Company | CONNECTED Chip Island
Semiconductor journalist Owen Hung-Wen Lin, who has covered TSMC since its early years, talks with host Divya Gopalan about why the chipmaker matters, its customer-first culture, its rivalry with Samsung, and Taiwan's geopolitical importance.
TED · Apr 3, 2026The Story You’re Not Hearing About AI Data Centers | Ayșe Coskun | TED
Computer scientist Ayșe Coskun describes how AI data centers are straining the power grid, and proposes running them as "virtual batteries" that help stabilize the grid and support clean energy. Recorded at TEDAI San Francisco in October 2025.
Veritasium · Dec 31, 2025The Closest Thing We Have to Alien Technology
Veritasium looks at the machines that make the most advanced computer chips: how microchips are made, how extreme ultraviolet (EUV) lithography works, and how ASML came to dominate chipmaking equipment, with a look inside the company.
ASML · Aug 21, 2025Inside the machine: How a High NA system generates EUV light | ASML
ASML's Greet Storms explains how the company's High NA EUV lithography machines make extreme ultraviolet light for chipmaking, by hitting tiny tin droplets to create a plasma that gives off 13.5-nanometer light about 50,000 times a second.
Made by the company itselfNVIDIA · Mar 18, 2025GTC March 2025 Keynote with NVIDIA CEO Jensen Huang
Nvidia CEO Jensen Huang presents the company's GTC 2025 announcements, including the Blackwell GPU architecture, Blackwell Ultra and Vera Rubin, the Dynamo inference software, and the idea of AI factories. The keynote also covers robotics, autonomous vehicles and new DGX systems.
Made by the company itselfAI and work.
What AI is doing to jobs, from economists, reporters and workers.
AI and your workUS payroll data shows young workers' jobs most affected by artificial intelligence • FRANCE 24
FRANCE 24 reports on new payroll figures showing employment of 22- to 25-year-olds in the most AI-exposed US jobs falling year on year, based on the Stanford Digital Economy Lab and ADP Research Canaries Dashboard, with an interview with ADP Research's chief economist.
McKinsey & Company · Aug 1, 2026The Biggest AI Opportunity Isn’t Replacing People | Stanford Economist
Stanford economist Erik Brynjolfsson discusses with McKinsey's talent team how AI is changing the career ladder, why productivity gains take time, and why redesigning work offers more than cutting headcount.
Faculty of Economics, University of Cambridge · Feb 25, 20262025-26 Stone Lecture - Professor David Autor
MIT economist David Autor gives Cambridge's 13th Sir Richard Stone Annual Lecture, "Expertise, Artificial Intelligence, and the Work of the Future," on how AI could change the value of human expertise at work.
MIT Sloan Management Review · Feb 23, 2026AI Is Not Improving Productivity: Nobel Laureate Daron Acemoglu
Nobel-winning economist Daron Acemoglu tells host Sam Ransbotham that the direction of AI is a choice, and argues for AI that creates new tasks for workers rather than mainly automating them. The conversation also covers claims about productivity, reliability risks and regulation.
Stanford Digital Economy Lab · Sep 30, 2025Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence
Stanford Digital Economy Lab researchers present their study of US payroll data, which found about a 13% relative drop in employment for 22- to 25-year-olds in the occupations most exposed to generative AI, while older workers and less exposed fields held steady or grew.
Pioneers of AI · Jan 30, 2025The future of work & education with AI (Wharton professor Ethan Mollick) | Pioneers of AI
Wharton professor Ethan Mollick talks with Rana el Kaliouby about using AI as a thought partner and treating it like a co-worker, and how AI is changing classrooms and workplaces.
Rules and regulation.
How governments are setting rules for AI, from the EU to the US and China.
Rules and regulationSenate holds hearing on White House's AI Action Plan | full video
White House science and technology adviser Michael Kratsios testifies before the Senate Commerce Committee about the Trump administration's AI Action Plan.
CNBC Television · May 8, 2025OpenAI CEO Sam Altman testifies on AI competition before Senate committee — 5/8/2025
The Senate Commerce Committee hears from Sam Altman of OpenAI, Lisa Su of AMD, Michael Intrator of CoreWeave and Brad Smith of Microsoft on strengthening US computing and innovation in the AI race.
Center for Strategic & International Studies · Apr 2, 2025Mapping Chinese AI Regulation with Matt Sheehan
Carnegie Endowment fellow Matt Sheehan explains how China's AI policymaking has developed over the past decade, which institutions shape it today, and how China's attitude to AI safety has changed.
Special Competitive Studies Project · Jun 19, 2024US and UK AI Safety Institutes, The First Year and the Road Ahead
Leaders from the US and UK AI Safety Institutes discuss the institutes' first year and plans ahead in a panel at the AI Expo for National Competitiveness.
European Parliament · Mar 12, 2024Europe’s AI act: first regulation of artificial intelligence
The European Parliament debates the Artificial Intelligence Act the day before its vote. The law, proposed by the European Commission in 2021, sorts AI systems by risk level.
Titles, channels and dates checked on YouTube Oct 25, 2026. Videos play in YouTube's own privacy-enhanced player; we don't copy or host them, and nobody pays to be listed. The pictures on the cards are our own drawings for each kind of video, not frames from the videos.