Skip to main content

ANDREJ KARPATHY 2025 LLM Review: Architectural Shifts, Vibe Coding, & The AGI Trajectory

[HPP] Andrej KarpathyJanuary 19, 20268 min
25 connections·36 entities in this video

The Rise of RLVR

  • 💡 Reinforcement Learning from Verifiable Rewards (RLVR) emerged as a fundamental fourth stage in LLM training, replacing RLHF.
  • 🎯 This new approach trains models on tasks with objectively verifiable answers, such as solving math problems or passing coding unit tests.
  • 🧠 RLVR forces models to discover complex multi-step problem-solving strategies independently, leading to what appears as
Knowledge graph36 entities · 25 connections

How they connect

An interactive map of every person, idea, and reference from this conversation. Hover to trace connections, click to explore.

Hover · drag to explore
36 entities
Chapters4 moments

Key Moments

Transcript30 segments

Full Transcript

Topics15 themes

What’s Discussed

Andrej KarpathyLarge Language Models (LLMs)RLVR (Reinforcement Learning from Verifiable Rewards)RLHF (Reinforcement Learning from Human Feedback)Jagged IntelligenceVibe CodingLLM AgentsLocal AgentsVisual GUIsMultimodal ModelsBenchmarksArtificial General Intelligence (AGI)Enterprise AdoptionSoftware DevelopmentTraining Pipeline
Smart Objects36 · 25 links
Person· 1
Concepts· 28
Media· 1
Company· 1
Products· 4
Event· 1