Have to give credit where it's due: Anthropic really turned the ship around with 5.5. Major improvements in prompt adherence over long running sessions.
Question is, will we see throttling over the next few days now that the "Anthropic is back" posts have proliferated?
Didn't think I'd be saying this so soon but Grok 4.7 is such a disaster that I'm back to using Opus for coding and copywriting.
I don't understand why 4.7 is so poor at prompt adherence and self checks (e.g. using agent judges to grade work). It's also near impossible to layer tasks because the initial set of rules is continuously forgotten and has to be restated.
@IndependentEco@proteahq I'm going to give it another week to see if they start limiting compute again, new models are always impressive the first week
Grok 4.7 is really not the improvement I was hoping for. Having a difficult time keeping it consistent across basic front-end tasks for @proteahq
I'm hesitant to switch back to claude as there's a fair chance they'll throttle inference again next week.
Maybe back to GPT...
Still amazed that it's now perfectly normal to move around town in a winged FSD vehicle without a steering wheel. The tranquility of the whole experience is just unmatched.
Was about to start a new session in Grok Build and noticed 4.7 is out. Seems to have further closed the gap to GPT and Claude. Putting it to the test right away.
Yep... their lineup is a mess, and fixing it doesn't seem to be a priority. Select tools are being halfway consolidated (e.g. moving some Grok bot features to Grok app but not all). Desktop app was supposed to be out months ago. Usage limits are inconsistent. I love the SpaceXAI ecosystem but this is a headache.
I think it's incredible that the greater tech community called this out weeks ago. Now it's mainstream, and confirms that open source poses as credible a threat to Sam and Dario's subscription enterprises as we thought.