Hacker News new | past | comments | ask | show | jobs | submit
it's the difference between a one-time transaction with a third party and an ongoing, perpetual third-party liability. you buy your inference hardware, you download your model, and if you wanted/needed to, you could run inference in perpetuity (or until your hardware broke down). compare that to anthropic/openai/etc being able to throttle intelligence, deprecate models, or raise prices on you whenever they felt like it.

sure, in a very general sense you're dealing with third parties in both scenarios, but if you actually own the hardware and model weights, you can control your own destiny for the lifespan of that hardware.

> it's the difference between a one-time transaction with a third party and an ongoing, perpetual third-party liability.

no ongoing liabilities except the power bill and opportunity cost of dropping $20k on hardware.

He isn't using OAI or Anthropic models. He is using models that area available for absurdly cheap from completely independent providers in multiple jurisdictions.

I get the base level argument, but from a practical point of view there are so many different dependencies that shifting one isn't really doing much.