The pages shown free above, from Local Ink Mock Title — 4/9/2026, as text — for screen readers and for searching within the issue. Automatically extracted, so spelling and layout may differ from the printed edition.
Page 1
Member-only story AI Just Solved an Open Problem in Theoretical Physics. And Nobody’s Talking About It A Google Research paper just dropped that quietly changes what we thought AI could do and it has nothing to do with chatbots. Rohit Kumar Thakur Following 7 min read · 1 hour ago 8 Photo by Bozhin Karaivanov on Unsplash I’ve been reading a lot of research papers lately. Most of them are incremental. Small improvements. Minor benchmarks. Nothing that makes you stop and think. But this one is a bit different. Researchers from Google Research, Harvard University, and Carnegie Mellon University just published a paper titled “Solving an Open Problem in Pub Crawl is happening this week! Sign up to join us!
Page 2
Theoretical Physics using AI-Assisted Discovery.” And when I say open problem, I mean a real one, a mathematical problem that human physicists had been trying to crack for years, couldn’t fully solve, and eventually just left on the table. Then an AI walked in and solved it. Let’s break this down. Just honest perspective on what happened, why it matters, and what it tells us about where we’re actually headed. First, What’s the Problem They Were Trying to Solve? Okay, I’ll keep this simple. There’s a concept in theoretical physics called cosmic strings . These are hypothetical, incredibly thin, incredibly dense structures in the universe, imagine a crack in the fabric of space-time left over from the Big Bang. Scientists believe these cosmic strings can vibrate and emit gravitational waves, ripples in space-time itself. To understand how much gravitational radiation a cosmic string emits, physicists need to calculate something called the power spectrum . And the mathematical engine behind that calculation is a complex integral, a mathematical formula that, when you try to solve it, runs into singularities (essentially, points where the math blows up and becomes unstable). Previous researchers had only managed partial solutions. They could handle certain special cases, like when the harmonic number is odd, or when they only looked at the behavior as N approaches infinity. But a clean, unified, exact solution for all cases? Nobody had pulled it off. That’s the open problem. And this is what the AI solved. So How Did AI Actually Do This? Here’s where it gets genuinely fascinating. The researchers didn’t just feed the problem into ChatGPT and ask for an answer. They built a sophisticated neuro-symbolic system, basically, they combined two things: the raw reasoning power of a large language model, and a structured, algorithmic search framework. The LLM they used was Gemini Deep Think , Google’s most advanced reasoning model. The search method they paired it with is called Tree Search (TS) . Imagine a branching tree where each branch represents a different way to solve the integral. Some branches work. Most don’t. The system’s job is to explore this tree smartly, cut off the paths that lead nowhere, and focus more on the paths that look promising.