AI dramatically lowers the cost of experimentation. Tasks that would be too tedious for a human, like rewriting an entire test suite to gauge performance impact, can be done by an agent in the background. This allows engineers to answer long-standing 'what if' questions almost instantly.

Standard AI benchmarks are an engineering tool for measuring performance. A more scientific approach, borrowed from cognitive psychology, uses targeted experiments. By designing problems where specific patterns of success and failure are diagnostic, researchers can uncover the underlying mechanisms and principles of an AI system, yielding deeper insights than a simple score.

The industry was surprised to learn that the tool-calling and problem-solving DNA of coding agents provides the necessary foundation for general-purpose agents. This was not the anticipated route to AGI, which labs hadn't explicitly trained for, yet it has become the dominant and most promising approach.

Just as biology deciphers the complex systems created by evolution, mechanistic interpretability seeks to understand the "how" inside neural networks. Instead of treating models as black boxes, it examines their internal parameters and activations to reverse-engineer how they work, moving beyond just measuring their external behavior.

As AI models are used for critical decisions in finance and law, black-box empirical testing will become insufficient. Mechanistic interpretability, which analyzes model weights to understand reasoning, is a bet that society and regulators will require explainable AI, making it a crucial future technology.

The ultimate goal isn't just modeling specific systems (like protein folding), but automating the entire scientific method. This involves AI generating hypotheses, choosing experiments, analyzing results, and updating a 'world model' of a domain, creating a continuous loop of discovery.

For AI systems to be adopted in scientific labs, they must be interpretable. Researchers need to understand the 'why' behind an AI's experimental plan to validate and trust the process, making interpretability a more critical feature than raw predictive power.

A new technique forces a model's forward pass to go through a natural language representation of its internal state. This makes the model's internal reasoning interpretable to humans in real-time, offering a significant breakthrough for monitoring and understanding what the model is actually "thinking" about a task.

Biohub applies mechanistic interpretability to its protein language models. By analyzing the model's internal representations—learned from both known and unknown biology—researchers can uncover emergent biological principles. This turns the model from a black box predictor into an engine for scientific discovery itself.