METHODS / STANDING RULES
These are the defaults behind every result on this site. Future posts link here rather than quietly changing the standard.
RESOLUTION
A difference smaller than the instrument’s variance is not a result. Repeats must match the boundary of the claim: within-session repeats for local comparisons, cross-session samples for claims that cross days or deployments.
PEDIGREE
Before interpreting an outcome, verify that the intended workload, configuration, and treatment reached the system. A green run proves completion—not experimental validity.
PREDICTION
When a model makes a testable prediction, record the range and assumptions before running the confirming experiment. A prediction written after the result is an explanation, not a test.
DISCLOSURE
Hardware, topology, generator model, dataset shape, run duration, exclusions, and known constraints travel with the claim. Caveats are part of the result, not footnotes to hide.
THE PRACTICAL STANDARD
Show enough work that a skeptical operator can reproduce the result, challenge the assumption, or decide that the measurement does not apply to their system.
I publish one measured post a month: get it by email.
No noise. Unsubscribe at any time.