AI models challenge one of Computer Science’s biggest assumptions

The result adds to growing anxiety in academia regarding the future role of human expertise

Published October 30, 2026
AI models challenge one of Computer Science’s biggest assumptions
AI models challenge one of Computer Science’s biggest assumptions

First it was Mathematics and now it is Computer Science. OpenAI and Anthropic both announced that their AI models have disproved the Exponential Time Hypothesis (ETH).

ETH is computer science’s foundational 25-year-old concept holding that certain complex computing problems can only be solved by testing nearly every potential solution.

When it comes to its significance, hundreds of academic studies in computer science and biology rely on this theory.

This claim marks the latest in a series of AI advancements that have disrupted academia, prompting profound reflection, especially among mathematicians, regarding the enduring relevance of human intellect and the trajectory of the field.

What is Exponential Time Hypothesis (ETH)?

Imagine you are going to solve a massive puzzle and you cannot solve this through a short-cut. It means you have to try every possible angle and combination to reach the result.

ETH claims that “to solve these hard problems, you basically have to try every single combination or close to it, which takes time that explodes exponentially."

Reactions from the hypothesis's key figures

The three researchers central to ETH, Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams who are helping organize an independent peer review of the AI companies' claims expressed their reactions over AI’s refutation.

Ryan Williams who is a longtime skeptic of ETH felt vindicated and views the event as part of a major revolution in the field.

On the other hand, Ramamohan Paturi from UC San Diego expressed surprise, while adding that existing research built around ETH will not disappear. He also emphasized that supremacy of human creativity remains central to guiding research.

Russell Impagliazzo’s reaction was based on unease, noting he would have preferred a human to discover the proof and explain their thought process.

He also raised concerns that these models could relegate humans primarily to verifying AI discoveries.

The researchers also sounded warning over the monopoly of tech giants over research tools. According to them, AI-generated results lack deeper explanations, impacting scientific understanding. 

Aqsa Qaddus Tahir
Aqsa Qaddus Tahir is a reporter dedicated to science coverage, exploring breakthroughs, emerging research, and innovation. Her work centres on making scientific developments understandable and relevant, presenting well-researched stories that connect complex ideas with everyday life in a clear, engaging, and informative manner.