Prismatic Agents

Methodology

The Evidence-First Comparison Method

We do not tell you what to choose. We organize the evidence, show where it came from, identify what remains unknown, and let you compare what matters to you.

Five evidence labels

Measured
We tested or measured it ourselves, with the method, hardware, date and limitations recorded.
Published
The maker or an authoritative first-party source publishes it. We link it and record when we checked.
Reported
A credible third party reports it. The wording reflects that it is secondary.
Estimate
Our own calculation. The formula, inputs and assumptions are always shown.
Not disclosed
No reliable public value was found. We never display it as “no”, “unsupported” or “does not have”.

What we never publish

We may show evidence completeness (how much is public), which measures information, not quality.

Our house rule on winners

Order, lenses and conflicts

The daily cycle

  1. 05:30 Toronto: every source is fetched and the values we track on it are compared with the last check.
  2. Unchanged sources re-confirm their facts; changed values go to a review queue.
  3. A Daily Brief is drafted. If nothing meaningful changed, it says so. We never manufacture daily stories.
  4. A person approves any changed value before it is published, website first, social after.

Independence and disclosure

Before anything AI-drafted is published