Blog
My random thoughts on science, math, and occassionally the places and animals I encounter.
Breaking ChatGPT-O1
OpenAI claims O1 is their most advanced model. I broke it on a simple math problem. Read More
Reproducibility in Biotech
Two labs have discovered a new mechanism in bacteria that allows them to synthesize a new gene from an RNA element. A gene that writes a gene. Wow. Read More
The central dogma, broken
Two labs have discovered a new mechanism in bacteria that allows them to synthesize a new gene from an RNA element. A gene that writes a gene. Wow. Read More
The cell cycle deserves all of your attention
I have a long-standing fascination with the cell cycle, because I think it's the key to understanding cell states, cellular differentiation and reprogramming. I assisted a conference that showed further evidence (in my opinion) for such a stance. Read More
Are AI models truly 'emergent'?
The most important term in AI today seems to be "emergence", and it refers to the point where a model "learns" something significant and not predictable. I argue that this is a dangerous concept, because it suggests that 'learning' is not quantifiable and therefore not predictable. Read More
For your genAI to work in biotech, pick the right modality
Resist the temptation to build an AI engine that can translate between all modalities. Instead, build an engine that translates all other modalities into the one that is most central to your hypothesis generation and testing approaches. Read More