# Vijay Pande thinks the era of massive > Source: > Published: 2026-08-29 17:44:53+00:00 # Vijay Pande thinks the era of massive The core of his argument rests on a massive paradigm shift: biology is transitioning from a "discovery" science into an "engineering" discipline. For decades, biology was about stumbling upon things—finding a protein that does X or a molecule that reacts with Y. It was observational and, frankly, quite messy. But with the integration of deep learning and massive compute, we are moving toward a world where we can design biological outcomes from the ground up. We aren't just looking for what exists; we are building what we need. However, this transition isn't without massive friction points that even the best LLM agents or protein-folding models can't solve overnight. Pande points out two massive hurdles that every AI-driven biotech startup will eventually hit: **The Clinical Trial Bottleneck:** Even if an AI can design a perfect drug candidate in seconds, the physical reality of human biology remains incredibly slow and brutally expensive. The regulatory and biological validation process is a hard ceiling that software alone cannot shatter.**Data Silos vs. Open Datasets:** This is perhaps the most critical technical debate in the industry right now. Many companies are trying to build "walled gardens" of proprietary data, thinking that owning a unique dataset is their ultimate moat. Pande argues the exact opposite. He believes that open, shared datasets are the actual catalyst that will allow AI to truly transform medicine. If we want to treat biology like an engineering problem, we need standardized, high-quality data that isn't locked behind a corporate firewall. In an engineering workflow, you need reliable, reproducible inputs. If every lab is hoarding its data, we lose the ability to train the foundational models required to understand the sheer scale of biological systems. This perspective is a reality check for anyone looking at the AI-biotech hype cycle. It isn't just about having a better prompt engineering strategy for protein design; it's about the underlying infrastructure of data and the physical reality of clinical validation. For those building in this space, the real winners won't necessarily be the ones with the most proprietary data, but the ones who can best leverage the collective intelligence of open-source biological datasets to accelerate the engineering loop. It's a high-stakes move from "let's see what happens" to "let's build this." [The DOJ investigation into a16z might change how VCs handle 7d ago](/en/news/7325/) [Next Leading open weights models are surprisingly easy to hijack with →](/en/news/8151/)