Scope
Beyond single-vendor dashboards
Provider-native consoles only tell part of the story. BitterBench is where local cells, cloud APIs, and routers become comparable.
Comparative Benchmark Product
Compare the same workload across the inference stack.
BitterBench is the product shell for comparable inference runs: sources, workload packets, run records, metrics, costs, and the evidence needed to decide which runtime actually wins.
Objects
Sources, workloads, runs, comparisons
The first shell names the comparative objects clearly so later gates can make them durable instead of inventing them mid-flight.
Phase
Product shell first
This scaffold is intentionally disciplined: real boundaries now, normalized benchmark truth next.