The success of iterative design processes hinges entirely on the metric being measured. An enzyme evolved for temperature stability won't necessarily remove clothing stains unless stain removal is the specific property being screened for. This highlights the critical importance of defining the right success metric from the start.
Related Insights
Foster a culture of experimentation by reframing failure. A test where the hypothesis is disproven is just as valuable as a 'win' because it provides crucial user insights. The program's success should be measured by the quantity of quality tests run, not the percentage of successful hypotheses.
Users mistakenly evaluate AI tools based on the quality of the first output. However, since 90% of the work is iterative, the superior tool is the one that handles a high volume of refinement prompts most effectively, not the one with the best initial result.
The traditional method of engineering enzymes by making precise, knowledge-based changes (“rational design”) is largely ineffective. Publication bias hides the vast number of failures, creating a false impression of success while cruder, high-volume methods like directed evolution prove superior.
The primary bottleneck in improving AI is no longer data or compute, but the creation of 'evals'—tests that measure a model's capabilities. These evals act as product requirement documents (PRDs) for researchers, defining what success looks like and guiding the training process.
With directed evolution, scientists find a mutated enzyme that works without knowing why. Even with the "answer"—the exact genetic changes—the complexity of protein interactions makes it incredibly difficult to reverse-engineer the underlying mechanism. The solution often precedes the understanding.
Frances Arnold, an engineer by training, reframed biological evolution as a powerful optimization algorithm. Instead of a purely biological concept, she saw it as a process for iterative design that could be harnessed in the lab to build new enzymes far more effectively than traditional methods.
Inspired by James Dyson, Koenigsegg embraces a radical commitment to differentiation: "it has to be different, even if it's worse." This principle forces teams to abandon incremental improvements and explore entirely new paths. While counterintuitive, this approach is a powerful tool for escaping local maxima and achieving genuine breakthroughs.
The best AI models are trained on data that reflects deep, subjective qualities—not just simple criteria. This "taste" is a key differentiator, influencing everything from code generation to creative writing, and is shaped by the values of the frontier lab.
Beyond optimizing existing biological functions, Frances Arnold's lab uses directed evolution to create enzymes for entirely new chemical reactions, like forming carbon-silicon bonds. This demonstrates that life's chemical toolkit is a small subset of what's possible, opening up a vast "non-natural" chemical universe.
The misconception that discovery slows down delivery is dangerous. Like stretching before a race prevents injury, proper, time-boxed discovery prevents building the wrong thing. This avoids costly code rewrites and iterative launches that miss the mark, ultimately speeding up the delivery of a successful product.
