The Sequence Opinion #868: Recursion Is the New Scaling Law

2026-05-28 · Source: TheSequence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

Modern AI progress, traditionally driven by scaling up models with more data and compute, is undergoing a significant shift. Historically, this "scaling law" approach defined the Transformer and foundation model eras, where larger models yielded better single-pass answers. However, recent advancements suggest a move away from this linear progression. The new frontier increasingly involves models and systems capable of iterative processes like revisiting, revising, searching, simulating, and critiquing their own outputs. This fundamental change redefines the important unit of computation from a single forward pass to a continuous loop, proposing that recursion, rather than mere size, is becoming the new scaling law for AI development.

Key takeaway

For AI Scientists and Directors of AI/ML evaluating future development strategies, recognize that simply increasing model size may no longer be the primary driver of progress. You should prioritize research and investment into systems that can iteratively refine, critique, and improve their own outputs. This shift towards recursive computation will define the next wave of AI advancements, demanding a re-evaluation of current scaling paradigms.

Key insights

AI progress is shifting from linear scaling to iterative, recursive processes for improvement.

Principles

• Traditional scaling focused on model size, data, compute.
• New frontier emphasizes iterative refinement and self-correction.
• Computation unit shifts from forward pass to loop.

Topics

• AI Scaling Laws
• Recursive AI
• Iterative Models
• Foundation Models
• Large Language Models
• AI Development Strategy

Best for: Research Scientist, AI Architect, AI Engineer, AI Scientist, Director of AI/ML, CTO

Related on AIssential

• The World Of LLMs Post Scaling Laws — AI model development has shifted from pretraining scale to post-training and verifiable data quality due to data exhaustion.
• Recursive Language Models: An All-in-One Deep Dive — RLMs use a REPL and pass-by-reference to enable LLMs to handle long contexts and complex tasks efficiently.
• BREAKING: Sam Altman concedes that we need major breakthroughs beyond mere scaling to get to AGI — Industry leaders are increasingly questioning pure scaling as the path to AGI, advocating for architectural breakthroughs instead.
• Reinforced Fast Weights with Next-Sequence Prediction — REFINE improves fast weight models for long-context tasks by shifting from next-token to next-sequence prediction via reinforcement learning.
• Generalization in LLM Problem Solving: The Case of the Shortest Path — LLMs generalize spatially but fail at length scaling in shortest-path problems due to recursive instability.
• Dario Amodei (Anthropic) Drops ATOMIC BOMBSHELL at Davos! — AI is rapidly approaching recursive self-improvement, driven by enhanced coding, mathematical intuition, and energetic efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by TheSequence.