The question boundary: what we ask machines, what we ask rebbeim

Every community runs on a question-routing table older than any device: the halacha shailah to the rov, the life crossroads to daas Torah, the chinuch puzzle to the mechanech who knows the child, the broken washing machine to the handyman. The table's genius is that it routes by the question's nature, not by whoever happens to be available — and its maintenance has always been a communal discipline, taught by example at a thousand kitchen tables. The conversational machine stresses the table in a new way: for the first time, there is an available-everywhere respondent willing to take every row — and fluent enough to sound, on each one, like it belongs there.

Why fluency is not authority

Name precisely what the machine lacks for the guidance rows, and the boundary draws itself:

• It does not know you. Daas Torah's whole mechanism is a wise person who knows the asker, the family, the history, the kehilla — guidance as a relationship's output. The machine's version is a composition for a generic asker who does not exist; the ask-versus-search distinction at its highest stakes.
• It carries no responsibility. The rov who rules carries the ruling — accountability before the community and Shamayim is part of the psak's substance. A machine's answer costs the machine nothing, which is exactly why it can afford to answer everything.
• It has no mesorah. The chain of transmission — who taught whom, which derech, whose shittos — is not decoration on Torah guidance; it is the guidance's provenance. Pattern-learned text has fluency without lineage, per the Torah-learning limits: a composition can echo the mesorah's words while standing entirely outside it.
• It optimizes for the asker's satisfaction. The machine's agreeableness — engineered, per the teen-companion risk — is the inverse of guidance: the rebbi who knows you sometimes tells you what you least want to hear, and that is the service. A respondent that never makes you uncomfortable is not guiding; it is mirroring.

“The rov's answer is expensive — a relationship, a wait, sometimes a no. The machine's answer is free and instant. The community's whole test is remembering that, for the questions that matter, the expense was the value.”

kolbo.life

The routing table, updated

The working table for a household with a guardrailed machine layer:

1. Tasks and explanations → the machine. The letter drafted, the concept explained, the plan structured, the homework tutored — composition's honest rows, where the machine serves brilliantly and nothing sacred is at stake.
2. Facts → retrieval. Sources, zmanim, the record — the search layer's rows, with addresses.
3. Halacha → the rov. The lookup-versus-shailah boundary unchanged: reference consulted, psak asked, and the machine's role at most a mareh-makom fetcher for the conversation you have with a human.
4. Life guidance → daas Torah and the people who know you. The shidduch decision, the chinuch crossroads, the parnassah fork, the emunah struggle — the rows where "who is answering?" is the answer. The machine's only kosher move here is the handoff — and a platform built for this community builds the handoff in: the KolBo AI posture of a machine that knows which rows it does not own.
5. The gray middle — thinking out loud. Using the machine to organize a hard question before bringing it to a person (the options listed, the considerations named) is legitimate and often excellent — provided the destination stays human. Preparation for guidance, never replacement of it.

Teaching the table

The boundary will hold exactly as well as households model it. The parent who says, audibly, "that's a good question for the machine" about the spreadsheet and "that one we ask the Rov" about the delicate matter is installing the routing table in every child within earshot — the same narrated-judgment chinuch as every other discernment the family builds. Worth making explicit at the table once: the community has never been anti-answer; it is pro-address. We are the people who ask "who says?" about everything — and a machine that answers everything is simply the newest reason that question was always the right one.

Frequently asked questions