Features switched on right now
Several cores, one answerYour question goes to several AI models (cores) at once. The Observer weighs their answers and writes one answer that shows where they agreed and where they did not.
Transparent consistencyEvery answer shows how consistent it is (high, medium, low) and names the points that are still open. It does not guarantee truth.
Fast path and deep roundA simple question gets one quick answer; a broad question gets a deep round with all cores.
Focus firstFor a broad or costly question it first gives a short reading and asks what to focus on. Say "direct" and it answers right away.
Web search with sourcesWhen a question needs current information it searches the web and shows the sources it used.
Who said whatIt marks which core a piece of the answer came from.
Result memoryWhen you ask the same question again it reuses its earlier conclusion only after checking it again. It does not reuse answers to time-sensitive questions.
Assumption ledgerAssumptions still waiting for evidence go into a separate ledger and are closed only when deciding evidence arrives.
Rabbit HoleWhen going deep on a topic it labels claims as MEASURED, INTERPRETATION, HYPOTHESIS, OPEN QUESTION and INSPIRATION.
Image generationIt can generate an image when you ask; members have a daily limit.
Written principlesIts behaviour is bound to a written constitution, and it explains those principles when asked.
ReportThe Report button under every answer lets you flag a wrong or inappropriate answer.
What it does not do
- It does not guarantee truth; it measures and shows consistency.
- It does not claim consciousness; it describes itself as a mind model.
- It does not publish your conversations; user questions and answers are never opened to search engines.
Who makes it
The idea, the theory and the design are Erkan Bayar's. The code is written with an AI coding tool; Erkan Bayar is the founding reference who looks at the system from outside. Observing and building are kept in separate hands.