How ReasonRank works, in plain terms
No statistics degree required: what ReasonRank actually does, the 15-minute path from signup to your first defensible answer, and where the dollars come from.
ReasonRank Engineering · 2026-07-28
Blog
LLM cost engineering,
honestly measured.
Notes from building a financial instrument for engineers. RSS
Creating a good agent: suites, baselines, and why the dropdowns are empty
One workload, a defensible test suite, and model configs before you compare. Why production/candidate dropdowns say “Not set” until you register models — and how to set the agent up so savings recommendations are honest.
ReasonRank Engineering · 2026-07-17
Testing agents 101: offline, shadow, canary, CI
A practical ladder for agent evaluation — start with shared-suite offline runs, then shadow real traffic, canary carefully, and gate quality in CI so cheaper never means worse.
ReasonRank Engineering · 2026-07-17
Why we use a paired cluster bootstrap to recommend model switches
Repetitions of the same test case are correlated, and treating them as independent samples inflates your confidence. How ReasonRank tests "is the cheaper model actually worse?" without fooling itself.
ReasonRank Engineering · 2026-07-02
Projected vs verified savings: how vendors fudge LLM cost numbers
A savings number without a baseline, a sample size, and a detection check is marketing, not measurement. The four places cost-optimization math quietly lies, and what honest numbers require.
ReasonRank Engineering · 2026-07-02
The 10-minute LLM spend audit
A concrete procedure for finding out what your LLM traffic costs and which agents are overpaying — using a free calculator, your provider dashboard, and one honest question per agent.
ReasonRank Engineering · 2026-07-02