Absence is information.
A target, estimate, internal run, or selected trace is not a public result. Until a repeatable observation is ready, the value remains N/A.
We publish the conditions around a result, not just the result itself. When the evidence is incomplete, the number stays blank. Ainfera separates intent from observation and leaves public values N/A until a repeatable record exists.
Read the public methodThe public sentence sits at the end of a chain. The model must be identified, the work must be defined, runtime conditions must be disclosed, and the observation must survive review. The figure below makes that threshold visible.
They are deliberately conservative. The cost of withholding a number is lower than the cost of publishing one without its meaning.
A target, estimate, internal run, or selected trace is not a public result. Until a repeatable observation is ready, the value remains N/A.
Model revision, peers, workload, hardware, runtime, quantization, and uncertainty define the comparison. Change the envelope and the claim changes with it.
A published value points back to the object, method, data, runtime, exclusions, and date required to inspect it. Context is part of the number.
Identity, protocol, runtime, and uncertainty determine whether two measurements belong in the same comparison. The claim boundary keeps those conditions visible.