Cowboy BioChemistry, Vibe-coding and the age of rapid iteration.

I was watching a video by Veritasium on protein folding, how AlphaFold 2 by Google DeepMind helped predict accurate protein structures from a bunch of amino acid data and was the only submission ever in the 14 editions of CASP to hit a score of 90.

Towards the end of the video, David Baker, a computational biologist, who won the Nobel Prize in 2024 for his work on computational protein design and for designing completely new proteins from scratch using RF Diffusion, an AI system developed by Baker’s Lab, calls his approach to building new protein structures as “Cowboy Chemistry”.

RF Diffusion is trained by adding random noise to a protein structure and then the AI has to remove that noise. After training the model this way, it can then be asked to produce proteins from scratch by providing random noise input and letting AI figure out a protein structure. They’ve already created human compatible anti-bodies that can neutralize lethal snake venom. His lab is now working on anti-bodies for auto-immune diseases.

The biggest take away for me was how rapidly David Baker and his team iterate. They have the designs on the computer, get the amino-acid sequence of the design proteins, and then in a couple of days get the protein out. David calls it “Cowboy BioChemistry” and in his words: “because you can just kind of go for it as fast as you can and it turns out to work pretty well.”

In the software programming world, vibe-coding is already a hit to the point where we have Andrew Ng launching a Vibe coding 101 course and close to 95% of the code for roughly 25% of the company’s in YC’s winter batch is LLM generated.=

We are only seeing the early examples of how wide the positive impact radius for large language models and AI today is. How it is prompting (no pun intended) leaders in other industries to use the technology to rapidly iterate, make transformative leaps and push forward the boundaries of human knowledge.