bench

Argument engine for qvra.

bench exists to create constrained comparison, reproducible measurement, and technical pressure. It is where claims get tested in public.

What belongs here

Artifacts in bench should provide one or more of:

• benchmark harnesses
• reproducible comparisons
• methodology notes
• challenger slots
• before/after evaluation
• reference measurements worth citing

What does not belong here

bench is not:

• marketing theater
• vague faster-than claims without method
• screenshots of results with no way to reproduce them
• prestige signaling through numbers alone

Anchor standard

Anything promoted into bench should satisfy these:

• explicit scope
• explicit comparison target or baseline
• explicit method
• outputs that can be checked
• a reason the result matters

Current role in qvra

bench is one of the gravitational anchors. It creates discussion, links, credibility, retesting, and technical argument under pressure.

Related routes:

• run for utility surfaces users can touch directly
• lab for experiments that may later deserve measurement
• show for visible demos of what is being measured
• pulse for the public rhythm of what changed

Publication rule

No benchmark belongs here unless the method is strong enough that disagreement can become productive instead of theatrical.

First live suite

• suites/line-count — reproducible comparison between wc -l and pure Python line counting on generated input

Status

Active anchor surface.