Proprietary data infrastructure builds a resilient, long-term competitive advantage.
The closed vs open model debate was about cost vis-à-vis performance.
Mythos just changed that. Now it’s about control. Renting frontier APIs works until the provider cuts you off.
Compelling argument for why owning your stack is the new edge.
http://x.com/i/article/2066399704347463680
Positive users note open-source AI can deliver superior accuracy via fine-tuning, while negative users emphasize unpriced control risks from relying on closed AI APIs.
Renting is fine until it isn't. The trap is you find out you needed to own the model the quarter the API gets capped or repriced, and by then the data work you skipped is a 12-month head start you don't have.
http://x.com/i/article/2066399704347463680
@dittycheria While your at it, drop a source-available document database into that stack! If AI is teaching us anything, flexibility to change is a core tenet.
Performance was not an issue in case of open source if we are speaking about accuracy. Open-source can deliver better accuracy because of fine-tuning, training, pre-training etc. in every enterprise use case (because of propriety data-set training and controlling complete experience)
@dittycheria Why dont you invest in india
@dittycheria enterprise software is indeed about control and accountability
@dittycheria Yup control risk is the real tax nobody's pricing in yet
