Real question, why are we forced to refactor for every model?

Even 3.5 had major changes to 4o and now system messages are changing again rendering countless apps useless to new models and forcing every developer else to refactor.

This seems to be in a vain effort to avoid DANs. Meanwhile DeepSeek doesn’t care and uses way less resources. Let’s talk about the elephant in the room.

Why does OpenAI feel obligated to fence in its users with a black box and “model Specs”? When DeepSeek so clearly shows its a waste of energy

2 Likes

AI is a rapidly evolving space. It is very easy to copy, but extreamly difficult to innovate and lead the way.

The price to pay in an evolving sector is change. I don’t think anyone wants to change things for the sake of change, it’s a required step in the evolution of AI inferencing.

I understand your frustration, but it comes with the territory, evey aspect of technology has change, AI is just more rapid than most.

2 Likes

I would like to interject for a moment and point out that a lot of these issues are OpenAI specific.

Yes, other vendors have issues too but OpenAI is creating issues out of thin air sometimes, for no apparent reason other than “safety”. That is very frustrating for a number of reasons.

I was going to mention that in their defense, some changes may have been necessary as more capabilities were added - but in retrospect that wasn’t exactly true.

2 Likes

I’ve had to move to Anthropic for one of my use cases because of too much “safety”. Shame. :frowning:

2 Likes