Bounded builder-time covers real deliverables, not slices, when the
work is Python-first: a scraper with proxy rotation and structured
output, an ETL pipeline that lands clean tables, a Django or FastAPI
backend with authentication, an automation that glues three APIs into
one workflow. What it does not cover in a single card is a migration
plus a rewrite plus a replatform, and the ceiling never pretends
otherwise; that is three cards, and the buyer ranks them. Slicing
ambition into card-sized deliverables is not a downgrade. It is the
difference between finished code landing in a private repo the buyer
owns every week and everything landing after a phase-gate review,
which is also the difference between a backlog that drains and one
that ossifies.