🎧 Prefer to listen?

Miles Wang, a researcher at OpenAI who dropped out of Harvard to work on AI-powered biological discovery, is reportedly in talks to launch his own startup — at a $2 billion valuation. Lightspeed Venture Partners is considering leading a $200 million round. And several other OpenAI researchers are expected to follow him out the door.

If you’re a solo builder watching the AI space, this story isn’t just biotech gossip. It’s a signal about where AI talent is moving, how the same transformer technology behind ChatGPT is being applied to entirely different industries, and what that means for the tools you’ll have access to in the next 12 months.

What Wang is actually building

The startup plans to use transformer architectures — the same fundamental technology that powers ChatGPT, Claude, and every other large language model — to identify new applications for existing drugs. Instead of spending $2.6 billion and a decade developing drugs from scratch, Wang’s approach focuses on repurposing compounds that already passed safety testing.

This is a smart play. The hardest part of drug development isn’t finding molecules that work — it’s proving they’re safe. Compounds that already cleared safety hurdles but failed for their original purpose still have potential. The challenge is figuring out what else they could treat. That’s exactly the kind of pattern-matching problem AI models are built for.

Wang co-authored research at OpenAI on using AI to automate and accelerate scientific discovery, and his proprietary approaches to applying transformers to biological data are apparently valuable enough to justify a $2 billion pre-money valuation before the company has a product.

Why this matters beyond biotech

The AI drug discovery sector has attracted over $15 billion in venture capital since 2024. Isomorphic Labs raised $2.1 billion. Xaira Therapeutics pulled in $1 billion. Chai Discovery closed $400 million. The pattern is clear: investors believe AI can fundamentally reshape how drugs get made.

But here’s what’s interesting for solo builders: the same technology Wang is using — transformer models applied to structured data — is becoming increasingly accessible through the tools you already use. Claude, ChatGPT, and Gemini are all built on transformer architectures. The difference between Wang’s startup and what you can do with these tools isn’t the underlying technology. It’s the training data and domain expertise.

We’ve covered this pattern before in our breakdown of how Anthropic’s Claude is entering science and pharma. The same AI models that help you write emails and automate workflows are being fine-tuned for drug discovery, materials science, and protein folding. The tools are converging. The applications are diverging.

The talent exodus pattern

Wang isn’t the first OpenAI researcher to leave for a high-profile startup, and he won’t be the last. This is what happens when a technology reaches a certain maturity level: the people who built it start seeing applications their employer isn’t pursuing.

For solo builders, this talent movement matters for two reasons:

First, it validates the “AI for X” thesis. When the people who built the foundational AI models start applying them to specific industries, it confirms that the technology is mature enough for vertical applications. You don’t see rocket scientists leaving NASA to build better toasters. You see them leaving when they realize the technology they developed can transform an entire industry that hasn’t been touched yet.

Second, it creates tool availability. Every time an OpenAI researcher leaves to build something specific, the broader AI ecosystem gets another data point about what’s possible. The techniques Wang developed will likely be described in papers, discussed at conferences, and eventually abstracted into tools that non-specialists can use. We’ve already seen this with AI agents becoming employees — what was cutting-edge research 18 months ago is now available as a service.

What solo builders should actually do with this information

Don’t try to build an AI drug discovery platform. That’s not the takeaway. The takeaway is this:

The transformer architecture is a general-purpose tool. The same attention mechanism that lets ChatGPT understand your prompt can be applied to molecular data, financial data, supply chain data, or any other structured information. Wang’s startup proves that the application layer — not the model layer — is where the value is being created.

If you’re building with AI tools, focus on the data, not the model. Wang’s $2 billion valuation isn’t because he has access to better transformers than anyone else. It’s because he has proprietary approaches to applying transformers to biological data, combined with domain expertise in drug development. The model is a commodity. The data and the domain knowledge are the moat.

Watch for tool democratization. The techniques that make AI-powered drug discovery work — molecular property prediction, compound-target interaction modeling, clinical trial outcome prediction — will eventually be packaged into tools that non-specialists can use. We’ve seen this pattern with AI coding agents, AI video generation, and AI-powered automation. Biotech is next.

The bigger picture

It’s not about Anthropic vs. OpenAI anymore. The AI industry is fracturing into verticals. The foundation model companies — OpenAI, Anthropic, Google — are becoming platforms. The real value is being created by people like Wang who take those platforms and apply them to specific, high-stakes problems.

For solo builders, this is both a threat and an opportunity. The threat: the AI tools you’re using today are being applied to solve problems in industries you might not have considered. The opportunity: the same tools that help you automate your business can be applied to problems in healthcare, finance, logistics, and dozens of other verticals. The question isn’t whether AI will transform your industry. It’s whether you’ll be the one doing the transforming.

The AI price war is making these tools cheaper. The AI groupthink problem is making it harder to see the real opportunities. And stories like Wang’s are showing where the smart money is betting.

Wang’s startup hasn’t closed its funding round yet. Details are still subject to change. But the signal is clear: the people who built the most powerful AI systems in the world are now applying them to the hardest problems in biology. And the tools they’re using are becoming available to everyone.

Start thinking about what your version of that looks like.

This post references reporting from TechCrunch, MLQ, and TechBuzz.