‘Mike often says this is a fascinating era—machines interrogating humans. What becomes possible when you can collect data from people on demand?

We could talk to every customer, not just the top 5% by contract value. We could proactively identify the root cause of every alert, instead of just reacting to problems.

The core mental model is: how would you build if you had an army of compliant, infinitely patient knowledge workers?

One important aside—many hard problems in the world don’t have answers in Common Crawl. The reasoning required for these problems isn’t captured in publicly available internet data.

These include fields like robotics, biology, materials science, and physics simulation. Solving them demands clever data collection and often interaction with atoms—not just bits. That can be intimidating for software people, but the payoff is worth it.

The same reasoning capabilities that can solve Math Olympiad problems can seemingly navigate molecular space. And if we work on these problems, I believe we can answer some of the most fundamental questions facing human society.’

– Sarah Guo at AI Engineer World’s fair 2025.