@FoxlightAI@TheTechPencil@ClaudeDevs My hypothesis is they removed the extra session limits that they introduced with Opus 4.7, so this is the true usage of 4.7.
@mocoydivad@DimitrisPapail If they force you to use the updated model, then from our perspective it is the same team member. And it creates a massive fulfillment headache as we have to spend time debugging and refinetuning processes instead of improving things.
@mocoydivad@DimitrisPapail More like "my team member just took a training course without asking me, and now he's impossible to manage and produces very inconsistent results which is resulting in more headachess"
@creativ_cyborg@AlexFinn Interestingly today (Saturday) it seems to be performing a little better (still not perfect). I read a post earlier that suggested during peak Silicone valley hours they dial down the reasoning tokens to allow for the increase in compute demand, so that would certainly explain it
@bcherny Oh, yay, more tokens to waste on the verbose dumbed-down model that can't think - it must be saving you more compute costs than you first expected, bravo! Bring back 4.6, this regression is breaking everything. I was about to buy another Max sub plan as well...
@louiseroddon@ObsMagazine Hi Louise, can you change the Striding Edge duration to something more conservative (4-6 hrs, depending on fitness), as well as mention that it is dangerous in snowy or icy conditions if the person lacks the right experience or equipment? Email me if you need more info.