Current AI investment rounds often function as low-yield subsidies for capital expenditures rather than backing groundbreaking innovation.

“here's a little secret when you put in a hundred million dollars into a startup to buy compute you are not buying whiz-bang Next Generation IP you are subsidizing capex and that is a job that many many other people do at a very low hurdle rate”

AI models are becoming a commodity, and future value will be driven by finding ways to deliver tasks, agents, and robotics rather than just chat interfaces.

“it's like these reasoning code agents computer use all this stuff is where opena has to actually go to make money in the future otherwise they're kaputs”