@thsottiaux@FintechKristen@ajambrosino Just to be clear is the reset on Tuesday going to be a global reset or a banked reset, I'd really appreciate some clarity on that
@thsottiaux I saw this somewhere else but since we have a fast mode will it be possible to create a slow mode? It will save us more token usage and it will save you'll more computing if most people end up using it
@thsottiaux Can you please address the token usage issue where it just seems to use 10% of the weekly limit on Sol medium standard speed even with simple prompts. If the 5 hour limit being gone is what's causing this please bring it back I guarantee most sensible people will agree
@thsottiaux Currently it feels like the weekly usage goes away very fast, I think that the 5 hour limit needs to come back because when we had the 5 hour limit the weekly limit never go as fast as it does now with small tasks
@thsottiaux I never switch to codex because there was never anything to switch from codex has and will always be the best option for me, as for what could be improved please look into gpt 5.6 Sol on high reasoning it feels like it uses alot more tokens that it needs to
@thsottiaux Just some feedback if the 5 hour limiter lowers token consumption alot than please can we bring it back. Because it feels like recently the token consumption has being going very quickly especially on gpt 5.6 Sol on high reasoning
@thsottiaux Codex is truly amazing and has helped me out alot with my projects as a result of how amazing it is we all tend to use up our weekly limit too quickly soo please can we have a reset π
@thsottiaux Please instead of global resets can we please make it always a banked reset? It's a bit unfair because we don't know when the reset is coming so we use a banked reset to get back to 100% and than 2 hours later we get a global reset. I do truly appreciate the resets and all tho
After using both extensively, Codex 5.6 Sol just feels much better than Fable 5 for real-world development. It understands large codebases better, follows existing architecture more consistently, and I spend far less time correcting its output.