Most of what we do day-to-day involves calling LLMs that already exist. But we wanted to know what’s actually happening underneath training, so we spent

After our first experiment on document-extraction, we ran a different kind of test on a different corpus. This one is a refusal question. We feed

A lot of our clients can’t send their data to a third-party API. Healthcare, finance, legal. Others can in principle, but don’t want a metered