AI Learned to Think Longer — and It Changed Everything
How AI learned to think longer — the second scaling law, inference-time compute, and the new era of reasoning models explained. You ask AI a question. It nails it. Change one word — it falls apart. Same model. Same knowledge. Completely different answer. Why? For over a decade, AI got better by getting bigger. More data. More parameters. More training compute. That was the only rule of LLM scaling. Then researchers noticed something strange. The knowledge was already inside the model — it just couldn't always find it. And when they let the model think longer at inference time, capability kept rising. Without new training. Without changing a single weight. This is the principle behind modern reasoning models like OpenAI o1, DeepSeek R1, and Claude's extended thinking. This is the story of test-time compute — the second scaling law — and why the next decade of artificial intelligence may look nothing like the last one. ━━━━━━━━━━━━━━━━━━━━━━━━━━━ 🧠 In this episode ━━━━━━━━━━━━━━━━━━━━━━━━━━━ • Why the same LLM gives different answers to the same question • How self-consistency broke the old rules of neural scaling laws • Chain of thought, generate-verify-select, and process reward models • Why search came back to AI — through reasoning, not games • Training compute vs inference compute: the two scaling laws • Where test-time compute stops working — and why that matters • What this means for o1, DeepSeek R1, and the next generation of AI ━━━━━━━━━━━━━━━━━━━━━━━━━━━ ⏱ Chapters ━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0:00 The Paradigm of Scaling 0:51 Self-Consistency 1:47 Parametric Search 2:17 Verifier Architectures 2:56 Inference-Time Scaling ━━━━━━━━━━━━━━━━━━━━━━━━━━━ 📡 The Latent Index ━━━━━━━━━━━━━━━━━━━━━━━━━━━ Finding signal in AI noise. Long-form explorations of how modern machine learning, large language models, and AI reasoning systems actually work — and where they're going next. 🔔 Subscribe so you don't miss Episode 2: MoE routing and speculative decoding — how AI decides which parts of itself to even use. ━━━━━━━━━━━━━━━━━━━━━━━━━━━ #ArtificialIntelligence #MachineLearning #AIExplained #LargeLanguageModels #ChainOfThought #ReasoningModels #ScalingLaws #InferenceCompute #o1 #TheLatentIndex

The Tiny Idea That Lets Anyone Fine-Tune AI

MIT Just Revealed the AI Bubble's Fatal Flaw

Stop Prompting Claude. Use Karpathy's Method Instead.

Why AI Can Never Escape Turing's 1936 Proof

Using Large Language Models | Build Your Own LLM Workshop #1

How To Think SO CLEARLY People Assume You're A Genius
![Yann LeCun's $1B Bet Against LLMs [Part 1]](https://i.ytimg.com/vi/kYkIdXwW2AE/hqdefault.jpg?sqp=-oaymwEjCNACELwBSFryq4qpAxUIARUAAAAAGAElAADIQj0AgKJDeAE=&rs=AOn4CLDbV4izF3i-wxevCVIn7FJjoy1vlA)
Yann LeCun's $1B Bet Against LLMs [Part 1]

LLMs Don't Need More Parameters. They Need Loops.

Psychology of People With Extremely High IQ

Rowan Atkinson's Brilliant Humor Leaves Celebrities in Tears!

Model Collapse Ends AI Hype

AI, Machine Learning, Deep Learning and Generative AI Explained

They're laughing at the SpaceX bubble

Anthropic is Completely F*cked.

Why Putin Is So Hard to Kill

Scientists Reveal Shocking Genetic Origin of Slavs

Billionaire's WARNING: I'm SELLING. The Crash Is Already Here!

This is not the AI we were promised | The Royal Society

AI Engineering in 76 Minutes (Complete Course/Speedrun!)

