🎧 Prefer to listen?

Miles Wang just left OpenAI to build a drug discovery startup at a $2 billion valuation. He’s not alone — several other OpenAI researchers are reportedly following him out the door. If you’re a solo builder, your first instinct might be to ignore this. Biotech is biotech, right? You’re building automations and landing pages, not running clinical trials. But that instinct is wrong — and I covered this exact pattern a few months ago when the trend first became visible.

Here’s what nobody is connecting: when the best AI researchers in the world leave general-purpose labs for specific industries, it changes the tools available to everyone. And what’s coming is going to make the tools you use today look primitive.

When researchers leave, the tools follow

This isn’t about one person switching jobs. It’s about a structural shift in where AI talent is concentrating. OpenAI researchers have historically worked on making better general-purpose models — ChatGPT, GPT-4, GPT-5, Codex. Their output was tools like ChatGPT, the OpenAI API, and Code Interpreter. Products designed for everyone, built by people who had never deeply engaged with any single domain.

The new batch of AI startups is different. Wang’s company isn’t building a general chatbot that happens to know about biology. It’s building models specifically trained on drug repurposing data — models that understand molecular interactions, trial outcomes, and FDA approval pathways at a level no general-purpose model can match. Chai Discovery raised $400 million at a $3.8 billion valuation doing something similar. Isomorphic Labs pulled in $2.1 billion.

The pattern is clear: general-purpose AI is hitting diminishing returns for business applications. The real value is in domain-specific models. And that creates a cascade effect that changes the tool landscape for every builder.

What changes for your workflow in the next 12 months

Think about the tools you use today. You probably have ChatGPT for writing and research, Claude for longer reasoning tasks, maybe Make or Zapier for automations, and Cursor or Bolt if you’re building apps. All of these are general-purpose. They work across domains because they’ve been trained on the entire internet — which sounds great until you realize they know a little about everything and a lot about nothing specific.

What’s coming is different. When AI researchers embed themselves in an industry for 18 months, they start building models that understand that industry’s data at a fundamental level. The tools that emerge from these labs won’t just be “ChatGPT with a pharma plugin.” They’ll be purpose-built systems that can do things general models simply can’t.

Here’s what that means in practice:

Domain-specific APIs are coming to every industry. Wang’s startup will almost certainly release APIs that pharmaceutical companies — and eventually smaller players — can use to screen drug candidates. But the pattern extends far beyond biotech. We’re already seeing it in legal AI tools that understand case law at a level ChatGPT can’t, and in financial AI models that parse earnings calls with context-aware precision.

For solo builders, this means your AI tool kit is about to get much more diverse. Instead of one general model doing everything, you’ll choose specialized models for specific tasks — the same way you already choose specific tools for specific jobs instead of using one app for everything.

The standard for “good enough AI work” is about to jump. Right now, you can impress clients with AI-generated copy, simple automations, and chatbot integrations. When domain-specific models become available, the bar moves. Your competitors will be offering AI that actually solves problems at an expert level. If you’re not tracking these developments, you’ll be using yesterday’s tools while the landscape shifts around you.

Regulated industries are the new frontier for tool builders. Media coverage focuses on the biotech angle, but the talent migration extends to finance, legal, insurance, energy, and agriculture. Every regulated industry is sitting on massive datasets that have never been properly modeled. The researchers leaving OpenAI aren’t just going to pharma — they’re spreading across every “unsexy” industry that used to be too complicated for tech startups. That creates opportunity for builders who understand how to use these tools.

The “college dropout founder” signal nobody’s talking about

There’s another detail buried in the TechCrunch story that’s worth paying attention to. Wang dropped out of Harvard to join OpenAI. And as the article notes, investors are now comfortable betting on founders who haven’t completed college — again.

If you’re a solo builder without a CS degree, this trend is in your favor. The value is increasingly in understanding how to apply AI tools to real problems, not in having credentials. Wang’s researchers will build the domain models. The builders who figure out how to connect those models to real-world use cases — through no-code tools, API integrations, and intelligent workflows — will capture the value.

Three moves to make right now

You don’t need to pivot into biotech. But you do need to track where AI talent is moving — because that’s where the next generation of tools comes from.

1. Watch the API landscape, not just the chatbot landscape. When domain-specific AI APIs start dropping — in pharma, legal, finance, agriculture — they’ll be the building blocks for your next client project. Subscribe to TechCrunch AI, follow AI tool launches, and test new APIs as they appear.

2. Build your workflow around model selection, not model loyalty. The era of “I use ChatGPT for everything” is ending. Start experimenting with Claude for research, specialized tools for specific domains, and open-source models for custom builds. The builders who know which tool to use for which job will have a massive advantage.

3. Position yourself in “boring” industries. While everyone else is building AI SaaS for other tech companies, the real money is flowing into healthcare finance supply chain compliance. These are industries where a 1% improvement translates to billions in value — and where solo builders who understand both the technology and the domain can charge enterprise rates.

The bottom line

The best AI researchers in the world are leaving general-purpose labs to solve specific, high-stakes problems in regulated industries. That’s not biotech news — it’s a signal about where the most powerful AI tools will come from next. The builders who pay attention now, learn to use specialized models, and position themselves in underserved industries will ride this wave. Everyone else will still be arguing about which chatbot is better.

Want help building your first AI workflow? Start here.