$SLC is in an interesting position as speech data starts becoming abundant. A million hours of speech data being made freely available is great for the industry, but it also exposes the limitation of measuring a dataset by hours alone. More data doesn't automatically mean better data. What matters is what sits behind those hours: • Data provenance • Recording quality • Language and script • Who contributed the recording • Consistency of the contributor • Accent and dialect representation • Whether they consented to its use This is where I think @silencionetwork is building something important. Silencio isn't just collecting more audio, it's creating mechanisms to understand the quality and provenance of the data being contributed. The new 5 star contributor ratings are a good example. A contributor's history and quality can become part of the dataset's signal instead of treating every recording as equally valuable. That matters because the speech data market is going to have no shortage of raw hours. The harder problem is finding clean, attributable, representative and trustworthy data that can actually be used. Free speech data makes volume cheaper and It makes quality more valuable. That's a pretty strong setup for $SLC .