I've worked on many migrations of things from vX to vX + 1, and there's always a tension between maximum backwards-compatibility, supporting every theoretical existing use-case, and just "flipping the switch" to move everyone to the New Way. Even though I, personally, am a "max backwards-compatibility" guy, it can be refreshing when someone decides to rip off the bandaid and force everyone to use the new best practice. How exciting! Unfortunately, this usually results in accidentally eliminating some feature that turns out to be Actually Important, a fuss is made, and the sudden forced migration is reverted after all.

I think the best approach is to move people to the newest version by default, but make it possible to use old versions, and then monitor switching rates and figure out what key features the new system is missing.

▲

ronsor 5 days ago | parent | next [-]

I usually think it's best to have both n and n - 1 versions for a limited time. As long as you always commit to removing the n - 1 version at a specified point in time, you don't get trapped in backward compatibility hell.

▲

koolala 5 days ago | parent [-]

Unless n is in any way objectively worse than n-1, then remove n-1 immediately so users don't directly compare them. Even Valve did it with Counter-Strike 2 and GO.

	▲	tibbar 5 days ago \| parent [-]
		With major redesigns, you often can’t directly compare the two versions —- they are different enough that you actually want people to use them in a different way. So it’s not that the new version is “worse”, it’s just different, and it’s possible that there are some workflows that are functionally impossible on the new version (you’d be surprised how easy it is to mess this up.)

▲

jillesvangurp 5 days ago | parent | prev | next [-]

These things have cost associated. In the case of AI models that cost comes in the form of massive amounts of GPU hardware. So, I can see the logic for OpenAI to not want a lot of users lingering on obsolete technology. It would be stupendously expensive to do that.

Probably what they'll do is get people on the new thing. And then push out a few releases to address some of the complaints.

	▲	tibbar 4 days ago \| parent [-]
		Are you saying that the hardware OpenAI used for inference on previous models is incompatible with the hardware used for GPT-5? Or are you perhaps saying that GPT-5 is just cheaper to run than the old models? > It would be stupendously expensive to do that. How are you quantifying this?

▲

djeastm 5 days ago | parent | prev [-]

>I think the best approach is to move people to the newest version by default, but make it possible to use old versions, and then monitor switching rates and figure out what key features the new system is missing.

See, one would think this would be the common sense approach and I thought was how they did it previously, no?

What's odd is that OpenAI didn't seem to feel it was worth doing this time around.