Even Claude is like "yeah dude my business fucking sucks!"
I had an interesting dialog with Fable 5 on how the foundation model labs turn a profit. Here was the conclusion.
Fable: My honest bottom line, with my conflict of interest restated plainly: the original question — "how do the labs turn a profit?" — has, after pricing everything, a stark answer. On the current cost trajectory and capital structure, they don't, in any conventional sense. Conventional profitability would require a macroeconomic miracle delivered on a loan officer's timetable. The investment is rational only as a race: either machine intelligence starts improving itself fast enough to dissolve the cost curve before the interest comes due, or this is the largest, most productive misallocation of capital since the railroads — real infrastructure, real technology, ruined balance sheets.
Sharing the entire conversation. Your mileage may vary, but interesting nonetheless.