Three problems that look separate are one: the app records money moving, and
separately records who owes whom, and the two never meet.
Documents what is broken with evidence - two half-built settlement models,
settlements existing twice unlinked, and Sonu's $37,980 of loan contributions
sitting unrecognised as generic transfers - then proposes settlement contexts,
payments as transactions rather than a side table, and loan co-ownership.
Nothing built. Five open questions, two of which are decisions about the
arrangement rather than the software.
siddharthd