On 26 October 2026, OpenAI said its AI models had produced three new scientific results: one in maths, one in computer science and one in physics. Not summaries. Not code tidy-ups. Original proofs of problems that top researchers have chewed on for decades.
If that holds up, it marks a shift. AI stops being just a clever assistant and starts doing the thinking behind discovery itself.
You don't need a physics degree to follow this. Below, we explain each claim in plain English, what it could mean for everyday life, and what it means for a small business like yours.
1. Maths: a better rulebook for how waves behave
The claim: OpenAI's models proved a long-standing maths problem about how waves spread out and bunch up over time (known as Sogge's local smoothing conjecture). As a knock-on, it says this also settles the famous Kakeya conjectures, which ask how much space you need to point a needle in every direction.
In plain English: Think of dropping a stone in a pond. Mathematicians have long known the ripples spread out, but not exactly how tightly they can focus into one spot. This proof gives a precise answer, in every dimension.
Why it matters day to day: Waves are everywhere. Sound, light, Wi-Fi, mobile signals, ultrasound and MRI scans all follow wave maths. Sharper theory tends to feed, years later, into:
- Clearer medical scans and imaging
- Better signal processing for phones, Wi-Fi and audio
- More accurate engineering models, from earthquakes to acoustics
You won't notice it next week. But this is the kind of foundation that quietly improves the tech you use in five to ten years.
2. Computer science: some "impossible" problems may be easier than we thought
The claim: OpenAI published a paper titled "The Exponential Time Hypothesis is False". For over 25 years, computer scientists assumed that certain puzzle-like problems could never be solved much faster than trying every option. OpenAI says its AI found a method that beats that limit, and checked it with Lean, a computer program that verifies proofs line by line.
In plain English: Imagine planning a wedding seating chart where dozens of guests can't sit together. Add more guests and the number of combinations explodes. The old belief was that no shortcut could ever tame that explosion. This result says a shortcut exists, at least in theory.
Why it matters day to day: These "tick every box" problems sit behind everyday decisions:
- Staff rotas and appointment booking
- Delivery routes and van scheduling
- Stock and supply planning
- Chip design and software testing
A reality check: The speed-up is mainly theoretical. It won't make your rota app faster overnight, and it doesn't break online banking or passwords. It also doesn't solve the famous P vs NP question. What it does is change researchers' map of what is possible, and that usually leads to better tools later.
3. Physics: putting the rulebook of the universe on solid ground
The claim: OpenAI says its models produced a rigorous mathematical proof that key theories of particle physics actually hold together. That includes quantum chromodynamics (the force that glues atoms' cores together) and a version of the Standard Model, physics' best description of matter. It also touches part of the Yang–Mills problem, one of the $1 million Millennium Prize puzzles, though OpenAI says it does not solve the prize in full.
In plain English: Physicists have used these theories for 50 years, and they predict experiments brilliantly. But nobody had proved, in strict maths terms, that they were watertight. It's like a bridge everyone drives over that has never had its full structural sign-off.
Why it matters day to day: The direct impact is long-term. A firmer foundation helps future work in:
- New materials and better batteries
- Medical physics, such as cancer radiotherapy and PET scans
- Energy research, including nuclear fusion
The bigger signal is this: if AI can do original physics, it can likely speed up research in drug discovery, materials and climate science too.
The catch: nothing is confirmed yet
These are claims, not settled science. OpenAI has handed the work to independent groups of leading academics, which it calls "academic shepherds", to check, explain and test each result. The computer science review, for example, is being organised by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams, two of whom first proposed the hypothesis in 1999.
The maths and computer science proofs come with computer-checked versions in Lean, which is a strong sign. But a machine check can't confirm that the problem was stated correctly. That takes human experts, and big claims like these often take months to review.
The sensible stance: be excited, not certain.
What this means for your business
You don't need to understand wave equations to benefit from this. The takeaway is simple: AI has moved from answering questions to solving hard problems. The same reasoning power is already available in the tools small businesses can use today.
Where UK SMEs are seeing real gains right now:
- Scheduling and rotas: AI that juggles staff, rooms and appointments in seconds
- Customer service: AI receptionists that answer calls and chats 24/7
- Admin and paperwork: drafting, summarising and filing on autopilot
- Decisions: turning your own data into clear, plain-English insight
The businesses that win won't be the ones with the biggest budgets. They'll be the ones that start small, test fast and build AI into everyday work.
Want to see what AI could do for your business? AI WORKLABS helps UK SMEs find practical, low-risk AI wins. Book a free chat