Source

. This is a substantive AI-bio / generative-science roadmap essay from an investor at Obvious Ventures, useful for the biotech-capital-cycle lane because it names the private-company layers and mechanisms that public-market proxies only imperfectly express.

The article's useful frame for our lane: the scarce asset in biology is not just compute or generic foundation-model capability. It is task-specific experimental data: assay outputs, binding curves, kinetic data, negative examples, proprietary wet-lab feedback loops, and data-generation factories that teach models the real constraints of biology.