Honest question, as I've just recently started fiddling with Meshtastic: could it be that the mesh is not set up correctly for a dense environment? (e.g. using LongFast rather than MediumFast, or not having more nodes configured as client_mute?) I know the conditions may be wildly different, but just as an example, the guy in this video says he saw no big issues on a hamvention with 300+ nodes: https://www.youtube.com/watch?v=aBfHAPpjtk4
My wife and I tried to use Briar to communicate after we had been reallocated two seat rows apart in a flight. It didn't work at all. Messages arrived hours later, when they arrived.
In a recent project I was asked to create a user story classifier to identify whether stories were "new development" or "maintenance of existing features". I tried both approaches, embeddings + cosine distance vs. directly asking a language model to classify the user story. The embeddings approach was, despite being fueled by the most powerful SOTA embedding model available, surprisingly worse than simply asking GPT 4.1 to give me the correct label.