The difference from an ordinary chatbot

Four differences you feel at work

• The source: every row here comes from a published register with an address, not from a model's memory.
• Repeatability: the same question returns the same rows today and in a month, with a hash to compare.
• The date: every layer carries a read date, so you know the age of an entry instead of guessing.
• The silence: if the row is missing, it is said outright. No row is invented to fill the gap.

What it does not do better

• It does not redraft a contract for you and does not compose a filing.
• It does not reason freely around an unregulated question.
• It does not replace judgement — it makes judgement cheaper to support.

FAQ

Can it invent a provision?

No. An answer consists of rows that exist in the register. If the row is not there you get no row, and the answer says so.

Why does a stable answer matter over time?

Because material that shifts between two occasions cannot be reviewed, and therefore does not work as documentation.

Should I stop using other assistants?

No. Use them where language is the task, and this one where the source is the task.