Скачать или смотреть How AI Agents Are Learning To Rewrite Themselves

Recursive AI is transforming artificial intelligence from a static tool into an evolving entity that can rewrite its own code. In this video, we explore the shift from "smart" models to "clever" agents that utilize recursive reasoning to outthink humans and break the autoregressive bottleneck.
Discover how the next generation of LLMs is moving beyond simple instruction following to master strategic depth and self-evolution.
In this video, you will learn:
The difference between AI being "Smart" (knowledge) vs. "Clever" (strategy).
How K-Level Reasoning allows AI to "mind-read" opponents in the Beauty Contest game.
How the Gödel Agent achieves recursive self-improvement by rewriting its own internal logic.
Why the future of AI depends on runtime evolution rather than just parameter scaling.
Timestamps:
00:00 Intro: Smart vs. Clever AI
01:27 The Mind-Reading Game: K-Level Reasoning
02:35 Why Humans Fail at Level 3 Thinking
03:15 The Gödel Agent: AI That Rewrites Its Code
04:18 Case Study: Solving the Game of 24
05:09 The Future is Recursive: Collective Intelligence
Sources:
Approximating Human Strategic Reasoning with LLM-Enhanced Recursive Reasoners
Design of a Self-Improving Gödel Agent with CrewAI and LangGraph
Gödel Agent: A Self-Referential Framework for Agents Recursively Self-Improvement
LADDER: Self-Improving LLMs Through Recursive Problem Decomposition
Recursive Language Models: the paradigm of 2026