TLDR Tech / Fintech
Ramp launches private SWE-Bench for financial coding agents
Ramp SWE-Bench is a private benchmark built from 80 production backend tasks that Ramp's internal coding agent, Inspect, previously shipped after engineer review. The benchmark evaluates how background coding agents handle financial software work across areas like card authorization, bill pay, accounting, procurement, treasury, and fraud. Pass/fail scoring is based on whether the agent produces a review-ready patch that flips tests from failing to passing within 45 minutes.
Read the full story