Awesome. OpenAI will eat this market.
Called it a couple weeks ago → https://news.ycombinator.com/item?id=41884821.
Some still think "OpenAI has no moat", keep on beating that horse, lol.
Called it a couple weeks ago → https://news.ycombinator.com/item?id=41884821.
Some still think "OpenAI has no moat", keep on beating that horse, lol.
It’s the same for all foundation model companies. The moat is how much money that have to sell inference and pay for training.
They'll be out of business very soon!
/s
You can stay in business for a long time with no moat and billions of dollars.
Long enough to find a moat even.
The proof of that is chatgpt itself. OpenAI did not pioneer a lot of those kinds of features. They saw startups doing it and just did it themselves since it’s fairly straightforward to implement.
A scrappy startup can copy the vast majority of chatgpt features with another foundation model.
Custom GPTs are just prompts.
Not for foundation models, and not when considered individually. But as long as OpenAI continues killing it with foundation models, continues moving up the stack with applications built on those models, and continues establishing market-leading parterships/integrations with companies like Microsoft and Apple, that's a moat.
True moats often involve factors like network effects, brand strength, intellectual property, or significant operational efficiencies that are harder to replicate.
"The combination of OpenAI’s advancements in foundation models, application development, and strategic partnerships collectively creates a robust moat for several reasons: [list of reasons]
"Together, these elements form a substantial moat. OpenAI’s cutting-edge models create a technological barrier, its applications generate a sticky user base, and its partnerships expand its influence and operational support, making it difficult for competitors to replicate its entire ecosystem. This combination of technological, user-based, and strategic advantages creates a robust, multi-layered moat around OpenAI in the competitive AI landscape."
Software is expensive to build, adoption is more expensive. If you ever try to convince someone to migrate in a B2B setting - you'll learn that its not enough to be cheaper, or better - you need to either be 10x cheaper or provide a next level experience that makes the existing product seem like a liability.
If you could copy ChatGPT's functionality (getting more expensive by the day), and offer it for $2/month - then folks might migrate. Otherwise the switching cost is too high for most.
You can. You can also make the core functionality of Twitter in a weekend.
Building software at this high a level isn’t too expensive or difficult, especially since the LLM provider you’d use handles the most difficult scaling issues for you.
> and offer it for $2/month
This you can’t really do. OpenAI can’t either tbf, they just have the runway to charge whatever they feel like… that cash is their moat.
I can upload CSV, JSON, PDF, any type of text file...
That's a UX implementation, not a moat. There's nothing ChatGPT-specific about vector stores/the Assistants API that Claude can't copy if there's enough demand.
I wonder whether running an ad on a ChatGPT conversation even pays for the cost of that conversation. I know the agent my company runs costs us 10x more to generate each response than an ad click would bring us in revenue. So, even if someone clicked on an ad every time they asked a question, we'd still lose money. Hmm.