Can a prompt replace Trello?

Build a kanban board and then do the two things that separate a good one from a demo: real drag-and-drop ordering, and a second view over the same cards.

Stack: your choice, Postgres, Docker Compose.

Boards, Lists, Cards. A Card has a title, description in Markdown, members, labels, a due date with a done flag, a checklist, comments, attachments and a cover colour.

Ordering is the part that is always done badly. Store position as a fractional value between neighbours rather than a dense integer sequence, so a drag writes one row instead of renumbering the list. Handle precision exhaustion by renormalising that list's positions in a background pass. Write a test that performs two hundred sequential drags to the same position and asserts the ordering stays correct.

Drag must be optimistic in the UI and reconciled against the server response, with a visible revert if the write fails — a board that silently loses a drag is worse than one that is slow.

Second and third views over the same cards, across all boards you are a member of:
- **Calendar**: cards with due dates, dragging a card changes the due date.
- **Table**: every card as a row with board, list, labels, members and due date, sortable and filterable, with inline editing.

Automation, kept small: rules of the form "when a card is moved to list X" or "when a due date passes" or "every weekday at time T", then actions from a closed set — move, archive, add or remove a label, add a member, post a comment, set a due date. Log every run. Cap chain depth at two.

Filters that persist per board per user: by label, member, due window, and free text.

Also: card templates per list, board archive with search, and JSON export of a whole board in a documented shape.

Write tests for fractional ordering under repeated inserts, for the calendar drag writing the same field as the card editor, and for automation chain-depth capping.

Do not build Power-Ups, mobile apps, or real-time multiplayer.