Learn AI by watching.
11 curated talks, tutorials, and keynotes — from Karpathy building GPT from scratch to the Google I/O 2026 announcements that defined the year.
11 videos
Andrej Karpathy
Intro to Large Language Models
A 1-hour general-purpose introduction to LLMs — what they are, how they're trained, and where the field is headed. Covers tokenization, fine-tuning, RLHF, and emergent capabilities without requiring a maths background.
Andrej Karpathy
Let's build GPT: from scratch, in code
Step-by-step implementation of a GPT language model in PyTorch, covering tokenization, self-attention, transformer blocks, and training on Shakespeare text. One of the most-watched AI coding tutorials on YouTube.
3Blue1Brown
But what is a neural network?
The definitive visual introduction to neural networks — how layers, weights, and activation functions combine to learn from data. Part of the acclaimed 'Neural Networks' series with stunning animations that make backpropagation intuitive.
Board
Mastering Data in FP&A
Deep dive into building analytical capabilities for Finance & Planning functions — covering data architecture, forecasting automation, and how AI tools are transforming the FP&A role from spreadsheet-driven to insight-driven.
3Blue1Brown
Attention in Transformers, visually explained
Visual explanation of the self-attention mechanism — the core innovation behind every modern LLM. Shows exactly how queries, keys, and values interact and why attention scales to long-context reasoning.
How to Power BI
AI Can Now Build ANY Power BI Chart You Want
Step-by-step walkthrough of using AI and Microsoft Copilot to generate any chart type in Power BI from a plain-English prompt — no manual drag-and-drop needed. Covers natural language chart generation, auto-written DAX measures, and how to prepare your data model so Copilot gives accurate results every time.
Andrej Karpathy
The spelled-out intro to neural networks and backpropagation
Part of the 'Neural Networks: Zero to Hero' series. Builds micrograd — a tiny autograd engine — from scratch to demystify how gradients flow through a network and how PyTorch's autograd works under the hood.
Google I/O 2026 Keynote
The full Google I/O 2026 keynote — Sundar Pichai and Demis Hassabis unveil Gemini 3.5 Flash, Gemini Omni (native video world model), Antigravity 2.0 agent platform, Gemini Spark (24/7 personal agent), AI Mode hitting 1 billion users, and Android XR audio glasses. The most product-dense I/O in Google's history.
Moonshots with Peter Diamandis
Anthropic Files $965B IPO, Trump Signs AI Executive Order, and ChatGPT Crosses 1B Users
EP #262 of the Moonshots podcast covers the week's biggest AI market moves: Anthropic's confidential S-1 IPO filing at a $965B valuation, Trump's AI executive order establishing voluntary 30-day pre-release reviews, ChatGPT crossing 1 billion monthly active users (the fastest consumer product adoption in history), Claude's 640% year-over-year growth, and OpenAI's push into robotics. Also covers Bernie Sanders' AI sovereign wealth fund bill and three major longevity breakthroughs.
Anthropic
Introducing Claude Fable 5
Anthropic's official launch video for Claude Fable 5, released June 9, 2026. Walks through the model's headline numbers — 95% on SWE-bench Verified, a 1M-token context window, and always-on adaptive thinking — and shows how it sustains long-horizon agentic tasks with built-in self-verification and sub-agent delegation.
Anthropic
Claude Fable 5 in Claude Code: The Hardest Coding Test Yet
A hands-on walkthrough of Claude Fable 5 running inside Claude Code on a multi-day refactoring task — using the memory tool to persist progress across sessions, delegating sub-tasks to background agents, and self-verifying its own changes before reporting completion. Demonstrates the kind of long-horizon, low-supervision workflow Fable 5's 1M-token context and adaptive thinking are designed for.