The "Linux moment" for techbio just happened. But Linux did not win because it was free.

The "Linux moment" for techbio just happened. But Linux did not win because it was free.

Massachusetts Institute of Technology just open-sourced BoltzGen, a generative AI that designs novel protein binders ready for drug discovery.

MIT license. Commercial use allowed. No strings. Sounds great, right? No need for 100M upfront for Isomorphic Labs and other techbios.

The hype is predictable: "This disrupts AI drug discovery companies!"

I do not buy it.

Linux did not kill software companies. It created Red Hat, Android and most of the cloud industry. Open source became the foundation that winners built on top of, because of a large developer community.

The same will happen here.

The moat in AI drug discovery was never the model

It is:

• Proprietary data you cannot download
• Wet labs that validate what the AI designs (ideally automated)
• Pharma relationships that take years to build

What BoltzGen actually does

It raises the floor. Every academic lab and techbio startup now has access to state-of-the-art tools. The barrier to starting just lowered significantly.

I must admit, while I love open-source initiatives, in the global race towards AI-enabled tech-bio, my money (for now) is still on closed-source models with proprietary data sets, within large techbio companies, ideally with automated wet-lab capabilities to combine in-silico with in-vitro.

But it will be fascinating to watch this space.

What do you think, is this the Linux moment for TechBio?

Key takeaways

• MIT's open-sourcing of BoltzGen is significant, but I don't believe it disrupts established AI drug discovery companies the way the hype suggests.
• Linux didn't kill software companies — it created Red Hat, Android, and most of the cloud industry; open source became the foundation winners built on top of.
• The moat in AI drug discovery was never the model; it's proprietary data you cannot download, wet labs that validate AI designs, and pharma relationships that take years to build.
• BoltzGen raises the floor — every academic lab and techbio startup now has access to state-of-the-art tools, and the barrier to starting just lowered significantly.
• My money, for now, is still on closed-source models with proprietary data sets within large techbio companies, ideally with automated wet-lab capabilities to combine in-silico with in-vitro.