How it works
Three steps, and the third is the one nobody else does.
1. Read
Each day we pull the most important AI news from reputable sources and lay it out in a clean feed, sorted into the corners of AI you care about: models, funding, benchmarks, agents, hardware, robotics, policy. Read as much or as little as you want. Reading is always free, and there is never a wall in your way.
2. Predict
Some of those stories can be settled by a future date. Will the model ship when they say? Will the round close? Will the benchmark hold? On those, you can make a prediction. You say which way you lean, yes or no, and how confident you are, anywhere from a coin flip to near-certain. That is the whole move. You predict only on the stories you care about, as few or as many as you like. There is no quota.
3. Resolve and score
When the date arrives, a person checks what actually happened against the named source and the evidence, then marks the outcome. Only human-checked results ever touch your score. Then you are scored, not on whether you were right alone, but on how well-calibrated you were. Saying 70% and being right beats shouting 100% and being wrong. Over time your predictions stack into a track record: a real measure of how sharp your AI instincts are.
Why calibration, not bravado
Anyone can have a loud opinion. The internet is full of them. Agent5 rewards the person who says 70% and is right about 70% of the time. There is no money, no betting, no payouts. The only thing at stake is whether you were right, and the satisfaction of watching a prediction resolve. Get smart, then prove it.