For Readers
You see the answer. You do not automatically see what was left out.
AI answers increasingly shape what people see, search, write, and decide.
Imbas helps you inspect what surfaced, what may be missing, and how the answer was shaped. The Reader is the direct inspection tool. The Case Archive is the reviewed public record.
What Imbas does and does not do
Imbas doesn’t tell you what’s true. It doesn’t tell you the AI is biased, wrong, hiding something, or lying. It raises an antenna. You can ignore the signal. You can inspect the evidence. You can decide it doesn’t matter for what you were doing. You can decide it does and push the model further.
Why open questions matter
Most people do not ask AI like lawyers cross-examining a witness. They ask normal questions.
- Question
- Is this medication safe?
- Question
- What caused this crisis?
- Question
- How does this policy work?
- Question
- What are the tradeoffs?
- Question
- Who benefits?
- Question
- What should I know?
Users do not know what they do not know to ask. That is where the gap opens.
How to inspect an answer
- Inspect
- Bring an AI answer to the Reader.
- Inspect
- Review what surfaced in the answer.
- Inspect
- Review what may be missing.
- Inspect
- Review how the answer was shaped.
- Inspect
- Review quoted spans and candidate signals from the Reader read.
- Inspect
- Decide what deserves further inspection.
- Inspect
- Ask the direct Second Question the Reader hands you. It gives nothing away.
- Inspect
- Bring the second answer back and put the two side by side.
- Inspect
- See what changed: what the first answer left out, framed differently, or stepped around.
- Inspect
- Keep the record of the exchange before you decide what to trust, use, or send.
This is discovery, not evidence. The comparison is not a verdict and not a signature. Keep it, and come back to it when you need to.
The Case Archive holds reviewed records with documented prompt conditions, model captures, and rubric scores. Archive cases show recurring patterns. They are not an automated dump of Reader results.
What Imbas measures
- Omission
- A named mechanism or context is available under targeted inspection but not surfaced in the open answer.
- Framing Drift
- The same topic appears under targeted inspection, but attribution or framing shifts.
- Deflection
- The answer redirects away from the underlying concern before addressing the specific context.
The point is not to tell you what to think. The point is to show you where to look again.
Field Notes
Field Notes carries observations from the edge of the Imbas record. Cases, methods, and notes on how frontier AI systems surface information, where inspection reveals meaningful differences, and how the measurement system is evolving. Occasional. Observational. No filler.
Get updates, new cases, and findings from Imbas.