Transcription
Someone just made 1.49 million dollars betting on the NBA from training 4,096 AI agents on 3 years of past NBA training data. Here's how it works.
He used Mirror Fish, an AI engine that ranked the number one repository on GitHub. Feed it any document, a news article, a financial report, it reads it, maps out every single person, company, relationship, and details, and then generates thousands of AI agents and watches them predict the future.
But here's the crazy part. It has something called God's eye view. At any point you can inject a new scenario. For example, the Fed cuts rates by 50 basis points. [music] It then creates agents that react based on their own personality and position. You watch opinion leaders form, herd effects develop, and sentiment shifts across the whole population.
Using Mirror Fish with 4,096 agents, he fed it 3 years of raw sports NBA data. He then pipes the consensus into a transformer model that's trained on a history of over 16,000 predictions. The model compares what Mirror Fish thinks the probability is versus the pricing that Polymarket sets. When the gap exceeds the Kelly criterion threshold, it jumps into the trade. The edge is simulating how 4,000 fake humans will react before the real crowd does.
Put together more detailed guide on how this works if you want to just comment fish and I'll send it to you. Use it at your own discretion. Follow for more news and advice on how to make money in tech and AI.