savas.one Back to home

Working experiments

Working AI experiments

Working counterparts to the ideas in the writing. Every experiment states what it demonstrates, which data it uses and where it refuses to answer.

Live · v1

LAB 01

AI Productization Lab

Compares how the same question is handled by a chatbot without evidence control and an AI product with evidence, scope and decision layers.

Compare live answers from the same SuperNova model with and without published savas.one evidence.

Open the live experiment →

QUERY

01Evidence
02Scope
03Control
04Decision

DECISION GROUNDED / REFUSE

Experiment design

Observable product behavior, not a showcase.

01

Evidence

The public source behind an answer is visible.

02

Refusal

Without evidence, the system does not invent a generic answer.

03

Measurement

Completed runs and decision types can be measured as anonymous events.