For most of history, life has been shaped by an invisible hand, sculpted by slow and imprecise forces. Over billions of years, random mutations and natural selection have given rise to the complexity we see today. This was the only mechanism available. But what if evolution were no longer bound by the sluggish pace of time? What if intelligence—artificial rather than biological—could rewrite nature’s playbook?

A new AI model, EvolutionaryScale Model 3 (ESM3), has demonstrated precisely this ability. It designed a functional protein that nature, left to its own devices, would have taken half a billion years to produce. With each advance in AI’s understanding of biological systems, the line between what is discovered and what is designed grows thinner. The consequences of this shift will be profound, not just for the field of molecular biology, but for how we understand the very concept of life.

The Intelligence Behind the Discovery

At its core, ESM3 is not unlike the AI systems that generate text, images, or code. Instead of language, it processes proteins—those fundamental molecules that drive every biological function. The model was trained on a vast dataset containing billions of protein sequences, millions of structural configurations, and extensive annotations describing their behaviors. With this knowledge, it can predict how proteins fold and what functions they may serve.

Instructed to create a glowing protein, much like the Green Fluorescent Protein (GFP) found in jellyfish, ESM3 generated something entirely new. The resulting molecule, esmGFP, was only 58% similar to naturally occurring GFP. In evolutionary terms, this represents an astronomical gap—one that suggests millions of years of hypothetical divergence. And yet, in the span of mere months, AI compressed eons of evolutionary trial and error into a single act of synthesis.

This isn’t just about efficiency. It represents a fundamental shift in our relationship with biology. Instead of observing and cataloging nature’s patterns, we are beginning to compose our own.

The Interwoven Threads of AI and Biology

AI has already transformed multiple domains, but biology remains uniquely poised for disruption. The logic of natural selection—variation, selection, and replication—mirrors many of the principles that drive AI models. What took evolution untold millennia to refine, AI can now iterate upon in real-time.

The implications are not limited to fluorescent proteins. AI-designed enzymes may soon break down plastic waste at a scale nature never evolved to handle. Entirely new materials could emerge—self-healing, bio-luminescent, or capable of energy storage. The ability to rationally design proteins opens doors that biology alone could never unlock. In medicine, custom proteins could precisely target cancer cells while leaving healthy tissue untouched. In agriculture, crops could be engineered to produce their own fertilizers. AI, trained on the vast library of nature’s solutions, is beginning to compose entirely new ones.

What Happens When Evolution Becomes Intentional?

The impact of this technology extends far beyond the immediate applications. The ability to engineer biology with such precision introduces new philosophical, economic, and ethical questions. If AI can generate proteins never before seen in nature, what does this mean for the future of evolution itself?

For most of history, humans have adapted to their environment. We are now reaching a point where we can redesign the environment to fit us. But when we modify life on this scale, are we solving problems or merely shifting their consequences elsewhere? A protein designed to break down waste could inadvertently disrupt ecosystems that have balanced themselves over millennia. A bioengineered crop resistant to disease could outcompete natural plants and alter entire food chains.

There is also the question of access. Who controls this technology? If AI-designed proteins unlock revolutionary medicines or industrial applications, will they be democratized or monopolized? The ability to engineer biology at will could become one of the most powerful forces of the coming decades, shaping economies and societies alike.

The Second and Third-Order Effects of an Engineered Biosphere

Some consequences of this shift are immediately obvious. Others will take time to reveal themselves. If AI can rapidly generate novel proteins, drug development could accelerate dramatically. In principle, we may enter an era where diseases are not merely treated but anticipated and countered before they arise.

Beyond medicine, entirely new industries may emerge around engineered materials. Natural selection never optimized for human needs. But what if AI did? We may soon see proteins designed to conduct electricity, store data, or sense environmental changes. The boundary between biology and technology will blur, and the concept of what it means to be “natural” may soon feel antiquated.

More subtly, as AI expands our capacity to manipulate biology, it will shift power in ways that are difficult to predict. If evolution itself becomes something that is designed rather than discovered, then what remains untouched by human intent? And if this technology is wielded unevenly, will we see new forms of inequality—those who can harness AI to reengineer biology and those who remain at its mercy?

Toward a Future of Designed Evolution

The creation of esmGFP is not an isolated event. It is a glimpse into a future where AI shapes biology as readily as it generates text or images. The forces that once dictated evolution—chance mutations, environmental pressures, and the slow churn of selection—are being supplanted by models that operate at an entirely different scale.

But the greatest transformation may not be technological. It may be conceptual. As we gain the ability to design new forms of life, we will be forced to reconsider the nature of life itself. What once seemed immutable—what it means to evolve, to adapt, to be alive—is now subject to revision.

This is not the end of evolution. It is its acceleration.