An AI model with “zero output tokens” was just open-sourced, yet the discussion volume on Binance Square surged to 9.33x first—are the technology real, or is the money real? Let’s look at the data first.
This round by FLock.io has rolled out a new model: single forward inference takes 31 milliseconds. It doesn’t use a decoder and doesn’t generate tokens—there are benchmarks shown in the paper too. It sounds pretty convincing.
But the paper also admits the shortfalls: multi-step arithmetic is weak, and the performance has some gloss within the training distribution.
Over the first 48 hours, projected discussion volume reached 15,909 posts, compared with a five-day average of 1,704—the timing was so tightly nailed that it’s hard to miss. The model was released from September 20 to 21, and the alert followed the very next day.
Leverage is the real face of this move: $25.31 million in open interest. Of the $265,000 liquidations, $149,000 was from short positions. Shorts were clearly hurt—this looks like a squeeze amplified by leverage, not new money flowing in. Also, this signal only counts Binance Square; whether other platforms like X followed is unknown.
I’m bearish on this move—I won’t chase it. The model is indeed a real improvement, but a market driven by discussion volume and leverage doesn’t prove the token can actually claim a slice of that “cake.” The only way for a turnaround is this: over the next few weeks, the discussion heat doesn’t fade, and once the shorts are cleared, price remains stable. Then—and only then—would it qualify as real funding taking over. With this data, it can’t yet.
$FLOCK #AI #Crypto
This round by FLock.io has rolled out a new model: single forward inference takes 31 milliseconds. It doesn’t use a decoder and doesn’t generate tokens—there are benchmarks shown in the paper too. It sounds pretty convincing.
But the paper also admits the shortfalls: multi-step arithmetic is weak, and the performance has some gloss within the training distribution.
Over the first 48 hours, projected discussion volume reached 15,909 posts, compared with a five-day average of 1,704—the timing was so tightly nailed that it’s hard to miss. The model was released from September 20 to 21, and the alert followed the very next day.
Leverage is the real face of this move: $25.31 million in open interest. Of the $265,000 liquidations, $149,000 was from short positions. Shorts were clearly hurt—this looks like a squeeze amplified by leverage, not new money flowing in. Also, this signal only counts Binance Square; whether other platforms like X followed is unknown.
I’m bearish on this move—I won’t chase it. The model is indeed a real improvement, but a market driven by discussion volume and leverage doesn’t prove the token can actually claim a slice of that “cake.” The only way for a turnaround is this: over the next few weeks, the discussion heat doesn’t fade, and once the shorts are cleared, price remains stable. Then—and only then—would it qualify as real funding taking over. With this data, it can’t yet.
$FLOCK #AI #Crypto