myTubehttps://pioed.com/2026-11-08T18:48:01.527ZThe spelled-out intro to neural networks and backpropagation: building microgradhttps://pioed.com/watch/VMj-3S1tku02026-11-08T00:00:00ZLecture 1 of Neural Networks: Zero to Hero, in which Karpathy builds a small autograd engine and neural network in Python to explain backpropagation step by step, and the Karpathy video the most Hacker News commenters recommend (17, from 2022 to 2025).AlphaGo - The Movie | Full award-winning documentaryhttps://pioed.com/watch/WXuK6gekU1Y2026-11-08T00:00:00ZGreg Kohs's feature documentary on DeepMind's 2016 match between AlphaGo and Go champion Lee Sedol, which premiered at the Tribeca Film Festival and has been recommended on Hacker News by 10 people since 2020.Deep Dive into LLMs like ChatGPThttps://pioed.com/watch/7xTGNNLPyMI2026-11-08T00:00:00ZKarpathy's three-and-a-half-hour general-audience lecture on the full training stack behind ChatGPT, from web data and tokenization through pretraining, supervised fine-tuning and reinforcement learning, recommended on Hacker News by 10 people in 2025 and 2026.[1hr Talk] Intro to Large Language Modelshttps://pioed.com/watch/zjkBMFhNj_g2026-11-08T00:00:00ZKarpathy's one-hour general-audience introduction to large language models, covering how they are trained, the LLM-as-operating-system analogy and jailbreak-style security problems, recommended on Hacker News by 9 people from 2023 to 2026.Let's build GPT: from scratch, in code, spelled out.https://pioed.com/watch/kCc8FmEb1nY2026-11-08T00:00:00ZLecture 7 of Zero to Hero, a GPT built from scratch following 'Attention Is All You Need' and GPT-2/GPT-3, recommended on Hacker News by 14 people and discussed in a 1,110-point thread.The Thinking Game | Full documentary | Tribeca Film Festival official selectionhttps://pioed.com/watch/d95J8yzvjbQ2026-11-08T00:00:00ZGreg Kohs's follow-up film, shot inside DeepMind over five years, following Demis Hassabis and the AlphaFold team to the protein-structure result that won the 2024 Nobel Prize in Chemistry.Andrej Karpathy: Software Is Changing (Again)https://pioed.com/watch/LCEmiRjPEtQ2026-11-08T00:00:00ZKarpathy's June 2025 keynote at Y Combinator's AI Startup School on Software 1.0, 2.0 and 3.0, LLMs as operating systems and partial-autonomy apps, which drew 1,481 points and 783 comments on Hacker News.How I use LLMshttps://pioed.com/watch/EWvNQjAaOHw2026-11-08T00:00:00ZKarpathy's February 2025 walkthrough of how he uses ChatGPT, Claude, Cursor, NotebookLM and other LLM tools, published as the third part of his general-audience series, so the product details date from that month.The spelled-out intro to language modeling: building makemorehttps://pioed.com/watch/PaCmpygFfXo2026-11-08T00:00:00ZLecture 2 of Zero to Hero, a character-level bigram language model built from scratch with PyTorch tensors, the starting point the later lectures grow into a GPT.Let's build the GPT Tokenizerhttps://pioed.com/watch/zduSFxRajkE2026-11-08T00:00:00ZLecture 8 of Zero to Hero, building the byte-pair-encoding tokenizer used by GPT models and tracing odd LLM behaviors back to tokenization, recommended on Hacker News by 5 people in 2024 and 2025.Building makemore Part 2: MLPhttps://pioed.com/watch/TCH_1BHY58I2026-11-08T00:00:00ZLecture 3 of Zero to Hero, which turns the bigram model into a multilayer perceptron and teaches training basics such as learning-rate tuning, train/dev/test splits and overfitting.Building makemore Part 3: Activations & Gradients, BatchNormhttps://pioed.com/watch/P6sfmUTpUmc2026-11-08T00:00:00ZLecture 4 of Zero to Hero, on how activations and gradients behave in deeper networks, how to diagnose them, and why initialization and batch normalization matter.Building makemore Part 4: Becoming a Backprop Ninjahttps://pioed.com/watch/q8SA3rM6ckI2026-11-08T00:00:00ZLecture 5 of Zero to Hero, in which Karpathy backpropagates by hand through a two-layer network with batch normalization without using PyTorch autograd.Building makemore Part 5: Building a WaveNethttps://pioed.com/watch/t3YJ5hKiMQ02026-11-08T00:00:00ZLecture 6 of Zero to Hero, which deepens the model into a tree-structured convolutional network similar to DeepMind's 2016 WaveNet.Andrej Karpathy - AI for Full-Self Driving at Teslahttps://pioed.com/watch/hx7BXih7zx82026-11-08T00:00:00ZKarpathy's ScaledML 2020 talk, given while he led AI at Tesla, on how Tesla trains vision neural networks from fleet data to replace hand-written driving code, recommended on Hacker News by 4 people from 2020 to 2022.State of GPT | BRK216HFShttps://pioed.com/watch/bZQun8Y4L2A2026-11-08T00:00:00ZKarpathy's 2023 Microsoft Build talk on how GPT assistants are trained, from tokenization and pretraining to supervised fine-tuning and RLHF, with prompting strategies, recommended on Hacker News by 6 people from 2023 to 2025.Let's reproduce GPT-2 (124M)https://pioed.com/watch/l8pRSuU81PU2026-11-08T00:00:00ZKarpathy's four-hour follow-on to the course, reproducing GPT-2 (124M) from an empty file, making training fast and running it overnight, which by his own description builds on the earlier Zero to Hero videos.Andrej Karpathy: Tesla AI, Self-Driving, Optimus, Aliens, and AGI | Lex Fridman Podcast #333https://pioed.com/watch/cdiD-9MMpb02026-11-08T00:00:00ZLex Fridman's 2022 conversation with Karpathy, recorded after he left Tesla, on camera-only self-driving, Optimus, transformers and AGI, recommended on Hacker News by 3 people from 2022 to 2025.Andrej Karpathy — “We’re summoning ghosts, not building animals”https://pioed.com/watch/lXUZvyajciY2026-11-08T00:00:00ZDwarkesh Patel's October 2025 interview with Karpathy on why agents will take about a decade to work, why he calls reinforcement learning terrible, why self-driving took so long and where education goes next, discussed in a 1,212-point Hacker News thread.Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AIhttps://pioed.com/watch/kwSVtQ7dziU2026-11-08T00:00:00ZSarah Guo's March 2026 No Priors conversation with Karpathy on coding agents, engineering and education, and his AutoResearch project in which agents design and run experiments without a human in the loop.How The Economic Machine Works by Ray Daliohttps://pioed.com/watch/PHe0bXAIuk02026-11-08T00:00:00ZRay Dalio's 31-minute animated model of how credit, debt and interest rates drive productivity growth and the short and long debt cycles, recommended on Hacker News by 18 people between 2014 and 2026.But what is a neural network? | Deep learning chapter 1https://pioed.com/watch/aircAruvnKk2026-11-08T00:00:00ZThe first chapter of 3Blue1Brown's deep learning series explains neurons, layers and weights by building a handwritten-digit recognizer, and 16 Hacker News commenters recommended it from 2017 to 2025 as the first stop for learning neural networks.“Hello, world” from scratch on a 6502 — Part 1https://pioed.com/watch/LnzuMJLZRdU2026-11-08T00:00:00ZBen Eater wires a bare 6502 processor on breadboards and watches its address and data lines as it starts up and reads its first instructions, the opening of the series 11 HN commenters recommend for learning how a computer works from first principles.Transformers, the tech behind LLMs | Deep Learning Chapter 5https://pioed.com/watch/wjZofJX0v4M2026-11-08T00:00:00ZChapter 5 follows text through a GPT-style transformer, from tokens and embeddings to the probabilities for the next word, and 13 Hacker News commenters recommended it from 2024 to 2026, many saying it made transformers click for them.Richard Feynman Computer Science Lecture - Hardware, Software and Heuristicshttps://pioed.com/watch/EKWGGDXe5MA2026-11-08T00:00:00ZRichard Feynman's lecture describing a computer as a very fast, very literal filing clerk, recommended by 8 HN commenters from 2013 to 2026 as a plain-language account of what a computer does; this copy is a reupload because no official upload exists.The Tragedy of systemdhttps://pioed.com/watch/o_AIw9bGogo2026-11-08T00:00:00ZFreeBSD developer Benno Rice traces how Unix systems have started and supervised their programs since the 1970s and why Linux moved to systemd, the operating-systems talk with the most HN recommenders in the HackerTube data.AT&T Archives: The UNIX Operating Systemhttps://pioed.com/watch/tc4ROCJYbm02026-11-08T00:00:00ZBell Labs' 1982 film in which Dennis Ritchie, Ken Thompson and Brian Kernighan explain the design of UNIX and demonstrate its shell and pipes, recommended by 10 HN commenters from 2013 to 2026 as a primary-source explanation of Unix.Gradient descent, how neural networks learn | Deep Learning Chapter 2https://pioed.com/watch/IHZwWFHWa-w2026-11-08T00:00:00ZChapter 2 shows how a network learns by scoring its mistakes with a cost function and nudging every weight downhill with gradient descent, and Hacker News commenters recommend the series as an accessible way into the math behind machine learning.William Ackman: Everything You Need to Know About Finance and Investing in Under an Hour | Big Thinkhttps://pioed.com/watch/WEDIj9JBTC82026-11-08T00:00:00ZPershing Square founder Bill Ackman builds a lemonade-stand company from nothing to explain raising equity and debt, reading a balance sheet and valuing a business before turning to investing, and Hacker News commenters point beginners to it as a crash course.Attention in transformers, step-by-step | Deep Learning Chapter 6https://pioed.com/watch/eMlx5fFNoYc2026-11-08T00:00:00ZChapter 6 works through attention, the mechanism that lets each word's vector take in context from the words around it, with queries, keys and values as concrete matrices, and Hacker News commenters point people who find text explanations hard to follow to it.The Thirty Million Line Problemhttps://pioed.com/watch/kZRE7HIO3vk2026-11-08T00:00:00ZCasey Muratori's lecture on why software now rests on tens of millions of lines of operating system and driver code, and on what a stable instruction set for whole chips would change, recommended by 11 HN commenters from 2018 to 2024.Geoffrey Hinton in conversation with Fei-Fei Li — Responsible AI developmenthttps://pioed.com/watch/QWWgr2rN45o2026-11-08T00:00:00ZA 2023 University of Toronto conversation in which Geoffrey Hinton and ImageNet creator Fei-Fei Li look back on the history of deep learning and discuss AI's risks, recommended on Hacker News by 2 people in 2024.Exploring How Computers Workhttps://pioed.com/watch/QZwneRb-zqA2026-11-08T00:00:00ZSebastian Lague builds logic gates in a simulator he wrote and uses them to represent and add binary numbers, showing in 18 minutes how on-off switches compute, and HN commenters recommend it as an introduction.Will AI outsmart human intelligence? - with 'Godfather of AI' Geoffrey Hintonhttps://pioed.com/watch/IkdziSLYzHw2026-11-08T00:00:00ZHinton's 2025 Royal Institution Discourse for a general audience on how neural networks learn, his argument that people use language much as trained networks do, and how digital intelligence could surpass ours, recommended on Hacker News by 3 people in 2025.USENIX ATC '21/OSDI '21 Joint Keynote Address-It's Time for Operating Systems to Rediscover Hardwarehttps://pioed.com/watch/36myc8wQhLo2026-11-08T00:00:00ZTimothy Roscoe of ETH Zurich argues in the joint USENIX ATC and OSDI 2021 keynote that a modern computer is a collection of processors and firmware that Linux only partly controls, a talk 12 HN commenters recommended from 2021 to 2026.13. Bankshttps://pioed.com/watch/1mDL1fKEVZM2026-11-08T00:00:00ZNobel laureate Robert Shiller's Yale lecture on why banks exist, why borrowing short and lending long exposes them to runs, and how deposit insurance and the Basel capital rules respond, from the Open Yale Courses Financial Markets course.Nobel Prize lecture: Demis Hassabis, Nobel Prize in Chemistry 2024https://pioed.com/watch/YtPaZsasmNA2026-11-08T00:00:00ZDemis Hassabis's Nobel lecture 'Accelerating scientific discovery with AI', given in Stockholm on 8 December 2024 for the AlphaFold protein-structure work the film above follows.The moment we stopped understanding AI [AlexNet]https://pioed.com/watch/UZDiGooFs542026-11-08T00:00:00ZWelch Labs opens up AlexNet, the 2012 image-recognition network, and shows what its layers learn to detect, from edges and colors in the first layer to more abstract features deeper in, using the feature-visualization and activation-atlas research it cites.19. Investment Bankshttps://pioed.com/watch/2yycGEFCNYE2026-11-08T00:00:00ZRobert Shiller's Yale lecture on what investment banks do (underwriting stock and bond issues, including IPOs), how their regulation moved from Glass-Steagall to Dodd-Frank and how the repo market spread the 2008 crisis, followed by a former student on moving from investment banking to Facebook.How do CPUs read machine code? — 6502 part 2https://pioed.com/watch/yl8vPW5hydQ2026-11-08T00:00:00ZIn the second part of the same series Ben Eater adds a memory chip, writes a program as raw machine code and follows the processor fetching and executing it, and HN commenters link it when pointing people to the series.Money, Power and Wall Street, Part One (full documentary) | FRONTLINEhttps://pioed.com/watch/W-Q9AOp2FW82026-11-08T00:00:00ZPart one of FRONTLINE's Emmy Award-winning 2012 series, in which correspondent Martin Smith interviews bankers, officials and journalists on how the big Wall Street banks grew, how they made their money and how that model led to the 2008 crash.Nobel Prize lecture: Geoffrey Hinton, Nobel Prize in Physicshttps://pioed.com/watch/XDE9DjpcSdI2026-11-08T00:00:00ZGeoffrey Hinton's Nobel lecture 'Boltzmann Machines', given in Stockholm on 8 December 2024, on the learning method behind his share of the 2024 Nobel Prize in Physics.Containers From Scratch • Liz Rice • GOTO 2018https://pioed.com/watch/8fi7uSYlOdc2026-11-08T00:00:00ZLiz Rice live-codes a container in a few lines of Go to show what Linux namespaces and control groups do and what a host knows about its containers, a talk 6 HN commenters recommend for understanding containers from the ground up.Linux Kernel Development, Greg Kroah-Hartman - Git Merge 2016https://pioed.com/watch/vyenmLqJQjs2026-11-08T00:00:00ZGreg Kroah-Hartman, who maintains the stable Linux kernel, explains how Linux is developed, how fast it changes and how its maintainers keep up, a talk HN commenters point newcomers to.The Origins of Linux—Linus Torvaldshttps://pioed.com/watch/WVTWCPoUt8w2026-11-08T00:00:00ZLinus Torvalds tells how Linux grew from his student project in Helsinki into a worldwide open-source kernel in this 2001 Computer History Museum talk, which HN commenters recommend watching in full.Jeff Dean & Noam Shazeer — 25 years at Google: from PageRank to AGIhttps://pioed.com/watch/v0gjI__RyCY2026-11-08T00:00:00ZTwo hours with Google chief scientist Jeff Dean and Noam Shazeer, co-inventor of the Transformer and mixture-of-experts models, on 25 years of building Google's systems from PageRank and MapReduce to Gemini.How to Understand and Choose a Venture Investorhttps://pioed.com/watch/VIOhMFw5DXY2026-11-08T00:00:00Za16z managing partner Scott Kupor, author of Secrets of Sand Hill Road, explains where venture funds get their money, what the limited partners behind them expect and why that pushes VCs toward companies that could become very large.How might LLMs store facts | Deep Learning Chapter 7https://pioed.com/watch/9-Jl0dxWQs82026-11-08T00:00:00ZChapter 7 covers the multilayer perceptron blocks that hold most of a transformer's parameters, works through a toy model of how they store a fact such as which sport an athlete plays, and explains superposition, citing the Google DeepMind and Anthropic research it draws on.CS162 Lecture 1: What is an Operating System?https://pioed.com/watch/pPzVV2kkGHc2026-11-08T00:00:00ZThe first lecture of Berkeley's operating systems course CS162, in which John Kubiatowicz explains what an operating system does and the context it runs in; teachyourselfcs.com names CS 162 as its operating-systems video course.AI can't cross this line and we don't know why.https://pioed.com/watch/5eqRuVp65eY2026-11-08T00:00:00ZWelch Labs explains neural scaling laws, the finding that a model's error falls along a predictable curve as compute, data and model size grow, walking through OpenAI's 2020 papers and DeepMind's Chinchilla result, and Hacker News commenters recommend it as a digestible introduction to the topic.