Inherent, founded by DeepMind alumni, says its AI 'teammate' just outperformed Anthropic and OpenAI at replicating research
The Specialist vs. The Generalist
While OpenAI and Anthropic focus on being everything to everyone, Inherent is betting on vertical depth. This is the classic struggle: can a specialized tool beat a Swiss Army knife at a specific, high-stakes job?
The Distribution Gap
Winning a benchmark is one thing, but getting a lab director to trust an agent with a multi-million dollar research pipeline is another. The winner won't just have the best model, they will have the most seamless integration into existing scientific workflows.
Beyond the Chatbox
We are seeing a shift from LLMs as conversational partners to LLMs as autonomous workers. If Faraday can truly replicate research, we enter the era of the agentic lab, where human scientists act more like directors than manual executors.
What to Watch
Look for Inherent to land a major enterprise partnership or release peer-reviewed validation of their results. If they can't prove it to the scientific community, it is just more hype.
Key Details
- Specialization is the best defense against Big Tech. If you can solve a high-stakes, narrow problem better than a general model, you have a real business.
- For builders, the win is not the model itself. It is solving the friction of getting AI outputs into actual lab equipment or research processes.
- Investors should look for proprietary data loops. If a startup's edge can be replicated by a better GPT-5 system prompt, it is a feature, not a company.