Our research used to sit inside Use Cases, next to case studies of how other companies use AI. It has nothing in common with them: it is work we do ourselves. So it now has its own section, Research, in the menu.
Its core question is a simple one: can an AI rebuild a real Georgian research report from public data? Each report is published with the prompts that made it and what it cost; where there is an official edition, ours goes out before it. Once the official report is out, we compare the two, number by number. The section’s list shows where each one stands: the result, or “waiting for the official report”.
Three releases so far: a test of five AI models, a weekly macro bulletin and a monthly employment tracker. A survey analysis follows on Saturday. Old links still work: they forward to the new addresses.