Great discussion, particularly on the future of software.

A lot of recent discussion has centered on unique data and workflows. After 20 years operating at Amazon and high-growth startups, Mike's point that humans possess sporadic knowledge that is surprisingly helpful resonated strongly.

This gets at the core problem: the interplay of hyper-personalized, compounding context and extremely dynamic reasoning at every turn.

The real unlock is capturing context and reasoning separately, then fusing them back together into a company-specific operating graph.

If you can combine the two while keeping each company's knowledge, data, company-specific workflows, and decisions private, you've created a new category of company.

Then you have to solve the incentive problem.

If you're selling automation tools to operators, they'll rarely be motivated to capture the full economic leverage. If you own the service across thousands of companies, you're fully incentivized to do so. Every improvement increases customer value, operator leverage, and gross margin simultaneously.

That's how an entirely new workforce emerges.

The one-person billion-dollar company is probably a transitional idea. The large outcome is a thousand-person AI workforce serving thousands of companies.

Every company will eventually have two execs: the CEO and the second executive. The CEO decides where to go. The second exec is the operating graph that executes.