Hallucination
When an AI states something false with complete confidence — an invented price, a policy you never had, a product you do not sell. The usual cause is that the correct fact was not available anywhere it could read.
angkor@design~query "Angkor Design"
4 matches — opening studio
Prefer to talk it through? Pick a time that suits you — free, no obligation. We'll meet on Google Meet or Telegram.Book a meeting →
When an AI states something false with complete confidence — an invented price, a policy you never had, a product you do not sell. The usual cause is that the correct fact was not available anywhere it could read.
Using historical data to estimate what will happen next: which customers will lapse, which stock will run out. Useful precisely to the degree that the past resembles the future.
Markup that labels questions and their answers as questions and answers. It is the most direct way to hand an assistant something it can quote in full without swallowing the surrounding page.
Splitting a long page into smaller passages so an AI can retrieve just the relevant part. Pages with clear headings chunk cleanly; a single unbroken wall of text does not.
How much text a model can hold in mind at once, counted in tokens. Anything beyond it is forgotten, which is why very long conversations start to lose earlier details.
The rough limit on how much of your site a crawler will fetch in a visit. Slow pages and endless near-duplicate URLs spend it on nothing, so the pages that matter get visited less often.
An extension to robots.txt that states not just who may crawl, but what they may do with what they find: appear in search, be quoted in answers, be used as training data. It lets you permit citation while refusing training.