We spent our first six months building double-entry accounting instead of features. I still don't know if that was right.
Here was the argument: anyone can generate a CRM now.
What you can't generate is the boring stuff underneath a real ledger, a payroll engine, a shared inbox that writes to the same records. So we built that first, and shipped the impressive-looking parts later.
The cost was real. We had nothing demo-able for months while people shipped AI tools every week.
For anyone who's made this call: did you build the unglamorous foundation first, or ship the visible thing and backfill? I'm curious whether the "boring first" bet actually pays off or just feels virtuous.