Frontier AI labs say their models refuted the Exponential Time Hypothesis. What does that mean?
News Source
β’Tue, 27 Oct 2026 17:08:08 +0000
π° What Happened
OpenAI says its AI models helped solve three major math problems. One result claims to disprove the Exponential Time Hypothesis, a long-held idea in computer science about which problems are truly hard to solve. The lab shared the work on Monday.
Researchers reacted with both excitement and worry. Anthropic mathematician Levent AlpΓΆge said OpenAI may have built on work that already existed. Hours later, Anthropic released its own version that also rejects the idea. OpenAI handled each field differently, giving mathematicians the work on GitHub while computer scientists formed a working group.
π The Backstory
The Exponential Time Hypothesis is an assumption, not a proven fact. It says some problems, such as checking every possible answer to a logic puzzle, cannot be solved much faster than brute force. For decades, researchers have used it to explain why certain tasks stay out of reach.
The idea is tied to the famous P versus NP question, which asks whether every problem that can be checked quickly can also be solved quickly. If the hypothesis falls, many results built on it need a second look. That is why a claim like this gets so much attention, and so much doubt.
π― Why It Matters
Proofs like this shape what scientists think computers can and cannot do. It also shows AI labs pushing into pure research, where mistakes are hard to spot and trust is thin.
OpenAI says its AI models helped solve three major math problems. One result claims to disprove the Exponential Time Hypothesis, a long-held idea in computer science about which problems are truly hard to solve. The lab shared the work on Monday.
Researchers reacted with both excitement and worry. Anthropic mathematician Levent AlpΓΆge said OpenAI may have built on work that already existed. Hours later, Anthropic released its own version that also rejects the idea. OpenAI handled each field differently, giving mathematicians the work on GitHub while computer scientists formed a working group.
The Exponential Time Hypothesis is an assumption, not a proven fact. It says some problems, such as checking every possible answer to a logic puzzle, cannot be solved much faster than brute force. For decades, researchers have used it to explain why certain tasks stay out of reach.
The idea is tied to the famous P versus NP question, which asks whether every problem that can be checked quickly can also be solved quickly. If the hypothesis falls, many results built on it need a second look. That is why a claim like this gets so much attention, and so much doubt.
Proofs like this shape what scientists think computers can and cannot do. It also shows AI labs pushing into pure research, where mistakes are hard to spot and trust is thin.