▲ | tosh 5 days ago | ||||||||||||||||||||||
would have been smart to keep them around for a while and just hide them (a bit like in the pro plan, but less hidden) and then phase them out over time would have reduced usage by 99% anyway now it all distracts from the gpt5 launch | |||||||||||||||||||||||
▲ | hinkley 5 days ago | parent | next [-] | ||||||||||||||||||||||
Charge more for LTS support. That’ll chase people onto your new systems. I’ve seen this play out badly before. It costs real money to keep engineers knowledgeable of what should rightfully be EOL systems. If you can make your laggard customers pay extra for that service, you can take care of those engineers. The reward for refactoring shitty code is supposed to be not having to deal with it anymore. If you have to continue dealing with it anyway, then you pay for every mistake for years even if you catch it early. You start shutting down the will for continuous improvement. The tech debt starts to accumulate because it can never be cleared, and trying to use makes maintenance five times more confusing. People start wanting more Waterfall design to try to keep errors from ever being released in the first place. It’s a mess. Make them pay for the privilege/hassle. | |||||||||||||||||||||||
| |||||||||||||||||||||||
▲ | Syntonicles 5 days ago | parent | prev [-] | ||||||||||||||||||||||
Is the new model significantly more efficient or something? Maybe using distillation? I haven't looked into it, I just heard the price is low. Personally I use/prefer 4o over 4.5 so I don't have high hopes for v5. |