Superposition

In April 2025, Anthropic CEO Dario Amodei published “The Urgency of Interpretability,” sounding an alarm that rippled through the AI research community. His message was stark: we’re building systems of unprecedented capability while remaining fundamentally unable to understand how they arrive at their outputs. The timing was deliberate—after years of incremental progress, a technique called Sparse Autoencoders (SAEs) had finally cracked open the black box, revealing millions of interpretable concepts hidden inside large language models. ...

In 1994, mathematician Peter Shor published an algorithm that would factor large integers exponentially faster than any known classical method. The cryptography community took notice—most of the world’s encrypted communications relied on the assumption that factoring large numbers was computationally intractable. Shor hadn’t built a quantum computer. He had merely proven that if one could be constructed, much of modern security infrastructure would crumble. Three decades later, quantum computers exist. They factor numbers, simulate molecules, and solve optimization problems. Yet they haven’t broken RSA encryption. The gap between having quantum computers and having useful quantum computers reveals something fundamental about the technology: quantum computing isn’t simply a faster version of classical computing. It’s an entirely different paradigm with its own physics, its own constraints, and its own challenges. ...