Humans created the hardest exam they could imagine to challenge AI... but AI is catching up at a terrifying pace
Humans created the hardest exam they could imagine to challenge AI... but AI is catching up at a terrifying pace
- Last year, the best AI scored just 8.8% on the world's toughest test
- This year it hit 38.3% — a 4x jump
- Some models now score over 50%, on questions designed by world-class experts
Ever feel like AI is getting smarter so fast it's scary? Yesterday it could barely do anything, today it's answering questions that stump experts? Wonder where it stops?
There's an exam called "Humanity's Last Exam" — created by top experts worldwide to test just how intelligent AI really is. The questions are so hard that even specialists in each field need deep, cutting-edge knowledge to answer them. Last year, the best AI managed just 8.8%. Many breathed a sigh of relief. But just one year later, that number leaped to 38.3% — and models like Claude and Gemini now score above 50%.
🎯 Why this matters:
- AI isn't improving gradually — it's improving in leaps
- The hardest exam humans could design is becoming solvable
- Expert-level knowledge is about to become accessible to everyone
- How we learn and work will transform dramatically within a few years
Think of AI as a student. Last year it failed nearly every subject. This year it passed almost half. If it were your child, you'd be proud. But if it were your competitor in a college entrance exam, you'd start to worry. The reality is that AI isn't a competitor — it's a tool. Those who learn to use it will have the advantage.
The world is entering an era where knowledge is no longer scarce. What's truly valuable now is the ability to ask the right questions and put the answers to good use. The age of "knowing a lot" is giving way to the age of "using it well."