As LLMs have practically exhausted all publicly available data, finding creative and insightful training data becomes a bottleneck for progress.
The issue with social media data like X is that it's biased by so-called influencers which the LLM (Grok) places more weight on. It's blindly following influencers who may or may not know right from wrong.
Through it's simple upvote/downvote system (and yes there is still some groupthink among certain subreds), Reddit filters our noise and allows novel ideas to percolate upwards.
In essence, Reddit is capturing the DNA of novel human thoughts.
So how soon will $RDDT be re-rated? LLM progress has plateau based on the current architecture, so the longer this is the case, the more valuable Reddit becomes.
I wouldn't be surprised if Reddit makes its data private and decides to build its own LLM at some point in the future.