A morning with the box

[EDITORIAL PLACEHOLDER — prototype scaffold, not the boxholder's words.] The narration below is demonstration text. The indented story cards are real: pulled verbatim, with provenance, from the verified user-story catalog in the repository. The emoji are stand-ins for character drawings.

Mail arrives

Overnight, three emails landed. At seven the box wakes up and pulls them in — each one becomes a card in the inbox, sitting next to everything else that needs attention.

As a box user, I want to pull incoming emails from Gmail into my box's inbox, so that I can process them locally as cards alongside my other work.

from beebox/docs/user-stories/catalog/2026-06-26.md
the bee turns an envelope over, unsure how it got inside the box

Nothing arrives by magic. A connector runs on each wakeup: it asks Gmail "what changed since last time," and writes what it finds as plain files — one card per thread, one per message. The box never sits inside your email; it copies what it needs and works locally.

generated

The box sorts

Most of the morning mail is routine, and routine things should not wait for a human. The box files what it recognizes and holds what it doesn't.

As a box operator, I want the agent to triage inbox items with three confidence levels (confident, probable, guess), moving them to category folders or holding uncertain ones with a review marker, so that I can quickly handle high-confidence items while reviewing edge cases.

from beebox/docs/user-stories/catalog/2026-06-26.md

A note on what “confidence” means here

It asks instead of guessing

One message fits nothing: an invoice from a company the box has never seen. So it stops, writes a question card, and moves on. The question waits on the dashboard until you answer it — and the answer teaches the box what to do next time.

why it asks instead of guessing

[PLACEHOLDER — the author's words go here.] This panel renders only the boxholder's own writing. Until that exists, it stays visibly empty — nothing here will be ghostwritten.

the author's words

Questions this page is asking

[Scaffold — reaction notes for the boxholder.] Do the three voices read as distinct at a glance, or do the kinds need more than an emoji and a provenance line (typography shift for the machine voice? a hand-drawn frame for the author?)? Is the bee's stage-direction label the right wordless register? Does quoting user stories as cards work as the walkthrough's spine, or should the spine be an actual transcript with the stories as asides? Is the provenance footer ("generated from the repository" / "the author's words") the right weight?