Open Models Turned AI Into Infrastructure
Dave Rogenmoser, CEO and co-founder of Jasper, on the generative AI opportunity
Stable Diffusion turned generative AI from a product you rented into a building block anyone could run, remix, and ship. OpenAI had already made text models easy to access through an API, but Stability made image generation feel like shared infrastructure. Developers could download weights, fine tune models, self host on their own GPUs, and build consumer apps without waiting for one central provider to approve access or set every price.
-
For app companies like Jasper, this changed the mental model of the stack. The model layer stopped looking like a scarce asset owned by one lab, and started looking more like cloud infrastructure that many startups could plug into, wrap with workflow software, and resell to users.
-
The floodgates point was not only more experimentation, but faster iteration. Open model weights meant developers and communities could fine tune, fork, and optimize quickly, while platforms like Hugging Face lowered distribution friction and made alternatives easy to find and compare.
-
That same openness also created the next competitive problem. When the underlying model is broadly available, value shifts upward into interface, workflow fit, customer data, and distribution. That is why Jasper pushed beyond simple copy generation toward embedded enterprise workflows and proprietary feedback loops.
Going forward, open models keep pushing the market toward an apps over models structure. Foundation model makers can still monetize through APIs and enterprise packages, but the durable winners at the application layer are the ones that own daily workflow, brand context, and user feedback, not just access to a model endpoint.