You should do a visualization (just a word cloud) of all your 100 ideas.
You *can* one up it by computing sentence embeddings and then visualizing all 100 ideas under t-SNE, but you might just get a boring distribution. Maybe contrast with others 100 ideas?
Or visualize the entire "conference" of 100-take threads :D gotta be a nice glitchy bit of art there (I was myself thinking of using the timestamps to create a dynamic view of the "Take Tsunami" wave unfolding over time
I think computing the embeddings for each of the opinions from each of the threads, and then overlapping the distributions in 2d t SNE but color coding so you can differentiate between threads would give the best result. You might find overlap between fields of thought.