At Tesla, they don’t run every possible simulation. They focus on the ones that produce real signal. The branches that generate learning. The boring ones get pruned because they teach nothing.
That idea kept rattling around in my head.
If we take the simulation hypothesis seriously (as he does), maybe the same applies more broadly: only timelines that produce learning are worth exploring. The ones where people take clear positions, make bold calls, and are willing to be wrong in public.
I built YourPrediction as a small experiment in that direction.
It’s a simple platform where people publish concrete predictions (for example, end-of-year takes), and then we track what actually happens over time.
There’s no algorithmic feed optimization or engagement tricks here, just predictions, outcomes, and discussion around why things played out the way they did.
At a minimum, it’s a tool to help us learn faster. At best, maybe it nudges us toward more interesting intellectual branches.
Happy to answer questions and iterate based on your thoughts.