34 chunks · 24 episodes
...models after multiple conversation turns now. This is good for everyone but the model providers; no individual model provider will have undue power by default, because there are multiple options in the same ballpark. Similar competitive dynamic...
...ion between the layers. The application layer and model layer will collide. The model providers will push hard to have their vertically integrated app used instead of the public API.
...main UI you use–the one that stores your context/memories–to also be one of the model providers.
...f features, because now users can get them for free… if they just commit to one model provider. It's OpenAI trying to change the game from stateless, easy-to-swap LLM providers, where the only competition is on quality and cost of the model, an...
...o be the dumb pipes. But everyone else wants the pipes they use to be dumb. The model providers don't want to be dumb pipes so they're moving aggressively up the stack to the application layer.
... it will take time to discover which ones. We are in the early innings! The LLM model providers are the electricity providers. Expensive, competitive, value-creating... but not necessarily a great business.
...t world! But less strategic power than the things that directly face users. LLM model providers will definitely be important... and also more likely to be subterranean. Unless the UX of actually using the models, e.g. high-quality integration w...
...s have a default advantage, but not a massive one. The reason it feels like the model providers' 1P UX will win is that it assumes that the "killer use case" of LLMs is a chatbot. If the killer use case is a vanilla chat bot, then it makes sens...
...l of the other data streams to feed LLMs, as evidenced by how willing the major model providers are to agree to contracts to not use any queries for training.
...memories) and is getting some stickiness… but it need not be the thing from the model providers. It's totally possible that someone could use off-the-shelf models from the top providers and create the sticky, value-generating service. The LLM p...
...your cookies and I will keep them safe" said the Cookie Monster. Interestingly, model providers don't seem to find the querystream particularly valuable and are willing to contractually give up the right to use it. In practice, we should all be...
...tickiness comes from becoming a paying subscriber. Indirect network effect: the model providers' actions imply the querystream isn't particularly valuable to increase model quality OpenAI is the kleenex of AI - if consumers know a single model ...
...rovider and owner of the data doing some unscrupulous. Even if one of the major model providers added Confidential Compute for their consumer app offering, it wouldn't change that much. The threat people worry about is the LLM provider training...
The main AI model providers are in an interesting strategic position. Their competitive differentiation is they're extremely good at producing models. All of them were research...