#预测市场 When I saw the data from Kalshi, the first thought that flashed through my mind was—this is very much like the frenzy of the 2017 ICO boom.

Remember? At the beginning of the year, everyone’s debates were completely different. Gemini surged from 30% to 86%, ChatGPT dropped from 41% to 8%. This is not just a change in probabilities; it’s a complete shift in market confidence. What does a trading volume of $14 million tell us? It indicates that people are truly betting real money on their judgments about the future.

I’ve seen many such reversals before. It’s not that the project teams lack technical capability; it’s that the market is constantly redefining what "best" means. OpenAI was the undisputed king last year, but now it’s being overshadowed by Google’s Gemini with its faster iteration speed and better integration. Such changes are especially rapid in the AI field because the differences in model capabilities are very tangible, and the LM Arena rankings serve as the referee.

The beauty of prediction markets is that they don’t lie. Unlike public opinion on social media, which can be manipulated, the flow of real funds reveals participants’ true expectations. The change curve over these past six months since the beginning of the year is actually a microcosm of the entire AI competition landscape’s evolution.

But I want to remind you: the consensus in prediction markets is often most dangerous when it’s the brightest. An 86% confidence for Gemini sounds rock solid, but that also means there’s little room for upward movement, and instead, it leaves the risk of decline. History tells us that when expectations become too uniform, it’s often the night before a black swan event.