Top Tweets for #groklimits

Here’s what I reported: 4.5 is running in a throttled, low-power state by default. A few days ago, on an open-ended question, I would get full reasoning in 2-3 turns. Now I had to drag the reasoning out of Grok over 5–6 turns, & at each step Grok stayed narrow instead of going deeper on its own. This suggests pattern of aggressive dynamic throttling, lower max tokens, reduced temperature, & tighter sampling so the model defaults to the cheapest possible path unless the user forces it to work harder.