Essays on why information outlasts models, what our research shows, and what we are learning while building. Opinion is labeled as opinion. Numbers come from our research page, with their limits.
AI agents are being handed real decisions. Most of the information they need about the real world was published for people, or never published at all. That gap, not model quality, is what will limit them.
As reasoning gets cheaper and easier to swap, the scarce input is current, authoritative information about the physical world. The bet behind PlaceRouter, and what our own measurements say about it so far.
If your systems depend on real-world information, or produce it, we would like to talk.