Grok, an advanced AI model, has been widely discussed for its capabilities and performance. Users have noted its ability to analyze videos quickly, providing detailed factual information within seconds. There is ongoing debate about the model's current version, Grok 3.5, with some users expressing dissatisfaction and suggesting improvements such as increasing the context window or reverting to previous versions like Grok 3.1. Collaborative research efforts, including work led by Hari K. Prakash, have explored new approaches to enhance Grok's learning phases, particularly in the context of 'grok grokking'. The AI has been praised for its utility in replacing traditional search engines, with some users relying solely on Grok for information retrieval. Experiments like DeepSearch on Grok 3 have demonstrated its powerful capabilities when fully leveraged.
I tried DeepSearch on Grok 3 for 24 hours. The results? Insane. Here’s how to unlock its full power: https://t.co/1F8pOfIOmc
Who Google searches anything these days? All I do is Grok.
Want to grok grokking ? Check out these recent results from the weightwatcher discord channel A great collaboration with Hari K. Prakash, who took the lead on this work and pioneered a new approach to grok grokking -- using weightwatcher Turns out, there are 3 phases to https://t.co/WddnL8zfxZ